From 16827862bcc4f04ada087abc255c6604d88076c1 Mon Sep 17 00:00:00 2001 From: "Wu, Ke" Date: Sat, 6 Dec 2014 12:17:27 -0500 Subject: Move non-MaxEnt code out of utils 1. alignment.h, argument_reorder_model.h, src_sentence.h, tree.h, tsuruoka_maxent.h -> decoder/ff_const_reorder_common.h. 2. Trainers source files (argument_reorder_model.cc and constituent_reorder_model.cc) are moved to training/const_reorder. --- decoder/ff_const_reorder_common.h | 1441 +++++++++++++++++++++++++++++++++++++ 1 file changed, 1441 insertions(+) create mode 100644 decoder/ff_const_reorder_common.h (limited to 'decoder/ff_const_reorder_common.h') diff --git a/decoder/ff_const_reorder_common.h b/decoder/ff_const_reorder_common.h new file mode 100644 index 00000000..7c111de3 --- /dev/null +++ b/decoder/ff_const_reorder_common.h @@ -0,0 +1,1441 @@ +#ifndef _FF_CONST_REORDER_COMMON_H +#define _FF_CONST_REORDER_COMMON_H + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "maxent.h" +#include "stringlib.h" + +namespace const_reorder { + +struct STreeItem { + STreeItem(const char *pszTerm) { + m_pszTerm = new char[strlen(pszTerm) + 1]; + strcpy(m_pszTerm, pszTerm); + + m_ptParent = NULL; + m_iBegin = -1; + m_iEnd = -1; + m_iHeadChild = -1; + m_iHeadWord = -1; + m_iBrotherIndex = -1; + } + ~STreeItem() { + delete[] m_pszTerm; + for (size_t i = 0; i < m_vecChildren.size(); i++) delete m_vecChildren[i]; + } + int fnAppend(STreeItem *ptChild) { + m_vecChildren.push_back(ptChild); + ptChild->m_iBrotherIndex = m_vecChildren.size() - 1; + ptChild->m_ptParent = this; + return m_vecChildren.size() - 1; + } + int fnGetChildrenNum() { return m_vecChildren.size(); } + + bool fnIsPreTerminal(void) { + int I; + if (this == NULL || m_vecChildren.size() == 0) return false; + + for (I = 0; I < m_vecChildren.size(); I++) + if (m_vecChildren[I]->m_vecChildren.size() > 0) return false; + + return true; + } + + public: + char *m_pszTerm; + + std::vector m_vecChildren; // children items + STreeItem *m_ptParent; // the parent item + + int m_iBegin; + int m_iEnd; // the node span words[m_iBegin, m_iEnd] + int m_iHeadChild; // the index of its head child + int m_iHeadWord; // the index of its head word + int m_iBrotherIndex; // the index in his brothers +}; + +struct SGetHeadWord { + typedef std::vector CVectorStr; + SGetHeadWord() {} + ~SGetHeadWord() {} + int fnGetHeadWord(char *pszCFGLeft, CVectorStr vectRight) { + // 0 indicating from right to left while 1 indicating from left to right + char szaHeadLists[201] = "0"; + + /* //head rules for Egnlish + if( strcmp( pszCFGLeft, "ADJP" ) == 0 ) + strcpy( szaHeadLists, "0NNS 0QP 0NN 0$ 0ADVP 0JJ 0VBN 0VBG 0ADJP + 0JJR 0NP 0JJS 0DT 0FW 0RBR 0RBS 0SBAR 0RB 0" ); + else if( strcmp( pszCFGLeft, "ADVP" ) == 0 ) + strcpy( szaHeadLists, "1RB 1RBR 1RBS 1FW 1ADVP 1TO 1CD 1JJR 1JJ 1IN + 1NP 1JJS 1NN 1" ); + else if( strcmp( pszCFGLeft, "CONJP" ) == 0 ) + strcpy( szaHeadLists, "1CC 1RB 1IN 1" ); + else if( strcmp( pszCFGLeft, "FRAG" ) == 0 ) + strcpy( szaHeadLists, "1" ); + else if( strcmp( pszCFGLeft, "INTJ" ) == 0 ) + strcpy( szaHeadLists, "0" ); + else if( strcmp( pszCFGLeft, "LST" ) == 0 ) + strcpy( szaHeadLists, "1LS 1: 1CLN 1" ); + else if( strcmp( pszCFGLeft, "NAC" ) == 0 ) + strcpy( szaHeadLists, "0NN 0NNS 0NNP 0NNPS 0NP 0NAC 0EX 0$ 0CD 0QP + 0PRP 0VBG 0JJ 0JJS 0JJR 0ADJP 0FW 0" ); + else if( strcmp( pszCFGLeft, "PP" ) == 0 ) + strcpy( szaHeadLists, "1IN 1TO 1VBG 1VBN 1RP 1FW 1" ); + else if( strcmp( pszCFGLeft, "PRN" ) == 0 ) + strcpy( szaHeadLists, "1" ); + else if( strcmp( pszCFGLeft, "PRT" ) == 0 ) + strcpy( szaHeadLists, "1RP 1" ); + else if( strcmp( pszCFGLeft, "QP" ) == 0 ) + strcpy( szaHeadLists, "0$ 0IN 0NNS 0NN 0JJ 0RB 0DT 0CD 0NCD 0QP 0JJR + 0JJS 0" ); + else if( strcmp( pszCFGLeft, "RRC" ) == 0 ) + strcpy( szaHeadLists, "1VP 1NP 1ADVP 1ADJP 1PP 1" ); + else if( strcmp( pszCFGLeft, "S" ) == 0 ) + strcpy( szaHeadLists, "0TO 0IN 0VP 0S 0SBAR 0ADJP 0UCP 0NP 0" ); + else if( strcmp( pszCFGLeft, "SBAR" ) == 0 ) + strcpy( szaHeadLists, "0WHNP 0WHPP 0WHADVP 0WHADJP 0IN 0DT 0S 0SQ + 0SINV 0SBAR 0FRAG 0" ); + else if( strcmp( pszCFGLeft, "SBARQ" ) == 0 ) + strcpy( szaHeadLists, "0SQ 0S 0SINV 0SBARQ 0FRAG 0" ); + else if( strcmp( pszCFGLeft, "SINV" ) == 0 ) + strcpy( szaHeadLists, "0VBZ 0VBD 0VBP 0VB 0MD 0VP 0S 0SINV 0ADJP 0NP + 0" ); + else if( strcmp( pszCFGLeft, "SQ" ) == 0 ) + strcpy( szaHeadLists, "0VBZ 0VBD 0VBP 0VB 0MD 0VP 0SQ 0" ); + else if( strcmp( pszCFGLeft, "UCP" ) == 0 ) + strcpy( szaHeadLists, "1" ); + else if( strcmp( pszCFGLeft, "VP" ) == 0 ) + strcpy( szaHeadLists, "0TO 0VBD 0VBN 0MD 0VBZ 0VB 0VBG 0VBP 0VP + 0ADJP 0NN 0NNS 0NP 0" ); + else if( strcmp( pszCFGLeft, "WHADJP" ) == 0 ) + strcpy( szaHeadLists, "0CC 0WRB 0JJ 0ADJP 0" ); + else if( strcmp( pszCFGLeft, "WHADVP" ) == 0 ) + strcpy( szaHeadLists, "1CC 1WRB 1" ); + else if( strcmp( pszCFGLeft, "WHNP" ) == 0 ) + strcpy( szaHeadLists, "0WDT 0WP 0WP$ 0WHADJP 0WHPP 0WHNP 0" ); + else if( strcmp( pszCFGLeft, "WHPP" ) == 0 ) + strcpy( szaHeadLists, "1IN 1TO FW 1" ); + else if( strcmp( pszCFGLeft, "NP" ) == 0 ) + strcpy( szaHeadLists, "0NN NNP NNS NNPS NX POS JJR 0NP 0$ ADJP PRN + 0CD 0JJ JJS RB QP 0" ); + */ + + if (strcmp(pszCFGLeft, "ADJP") == 0) + strcpy(szaHeadLists, "0ADJP JJ 0AD NN CS 0"); + else if (strcmp(pszCFGLeft, "ADVP") == 0) + strcpy(szaHeadLists, "0ADVP AD 0"); + else if (strcmp(pszCFGLeft, "CLP") == 0) + strcpy(szaHeadLists, "0CLP M 0"); + else if (strcmp(pszCFGLeft, "CP") == 0) + strcpy(szaHeadLists, "0DEC SP 1ADVP CS 0CP IP 0"); + else if (strcmp(pszCFGLeft, "DNP") == 0) + strcpy(szaHeadLists, "0DNP DEG 0DEC 0"); + else if (strcmp(pszCFGLeft, "DVP") == 0) + strcpy(szaHeadLists, "0DVP DEV 0"); + else if (strcmp(pszCFGLeft, "DP") == 0) + strcpy(szaHeadLists, "1DP DT 1"); + else if (strcmp(pszCFGLeft, "FRAG") == 0) + strcpy(szaHeadLists, "0VV NR NN 0"); + else if (strcmp(pszCFGLeft, "INTJ") == 0) + strcpy(szaHeadLists, "0INTJ IJ 0"); + else if (strcmp(pszCFGLeft, "LST") == 0) + strcpy(szaHeadLists, "1LST CD OD 1"); + else if (strcmp(pszCFGLeft, "IP") == 0) + strcpy(szaHeadLists, "0IP VP 0VV 0"); + // strcpy( szaHeadLists, "0VP 0VV 1IP 0" ); + else if (strcmp(pszCFGLeft, "LCP") == 0) + strcpy(szaHeadLists, "0LCP LC 0"); + else if (strcmp(pszCFGLeft, "NP") == 0) + strcpy(szaHeadLists, "0NP NN NT NR QP 0"); + else if (strcmp(pszCFGLeft, "PP") == 0) + strcpy(szaHeadLists, "1PP P 1"); + else if (strcmp(pszCFGLeft, "PRN") == 0) + strcpy(szaHeadLists, "0 NP IP VP NT NR NN 0"); + else if (strcmp(pszCFGLeft, "QP") == 0) + strcpy(szaHeadLists, "0QP CLP CD OD 0"); + else if (strcmp(pszCFGLeft, "VP") == 0) + strcpy(szaHeadLists, "1VP VA VC VE VV BA LB VCD VSB VRD VNV VCP 1"); + else if (strcmp(pszCFGLeft, "VCD") == 0) + strcpy(szaHeadLists, "0VCD VV VA VC VE 0"); + if (strcmp(pszCFGLeft, "VRD") == 0) + strcpy(szaHeadLists, "0VRD VV VA VC VE 0"); + else if (strcmp(pszCFGLeft, "VSB") == 0) + strcpy(szaHeadLists, "0VSB VV VA VC VE 0"); + else if (strcmp(pszCFGLeft, "VCP") == 0) + strcpy(szaHeadLists, "0VCP VV VA VC VE 0"); + else if (strcmp(pszCFGLeft, "VNV") == 0) + strcpy(szaHeadLists, "0VNV VV VA VC VE 0"); + else if (strcmp(pszCFGLeft, "VPT") == 0) + strcpy(szaHeadLists, "0VNV VV VA VC VE 0"); + else if (strcmp(pszCFGLeft, "UCP") == 0) + strcpy(szaHeadLists, "0"); + else if (strcmp(pszCFGLeft, "WHNP") == 0) + strcpy(szaHeadLists, "0WHNP NP NN NT NR QP 0"); + else if (strcmp(pszCFGLeft, "WHPP") == 0) + strcpy(szaHeadLists, "1WHPP PP P 1"); + + /* //head rules for GENIA corpus + if( strcmp( pszCFGLeft, "ADJP" ) == 0 ) + strcpy( szaHeadLists, "0NNS 0QP 0NN 0$ 0ADVP 0JJ 0VBN 0VBG 0ADJP + 0JJR 0NP 0JJS 0DT 0FW 0RBR 0RBS 0SBAR 0RB 0" ); + else if( strcmp( pszCFGLeft, "ADVP" ) == 0 ) + strcpy( szaHeadLists, "1RB 1RBR 1RBS 1FW 1ADVP 1TO 1CD 1JJR 1JJ 1IN + 1NP 1JJS 1NN 1" ); + else if( strcmp( pszCFGLeft, "CONJP" ) == 0 ) + strcpy( szaHeadLists, "1CC 1RB 1IN 1" ); + else if( strcmp( pszCFGLeft, "FRAG" ) == 0 ) + strcpy( szaHeadLists, "1" ); + else if( strcmp( pszCFGLeft, "INTJ" ) == 0 ) + strcpy( szaHeadLists, "0" ); + else if( strcmp( pszCFGLeft, "LST" ) == 0 ) + strcpy( szaHeadLists, "1LS 1: 1CLN 1" ); + else if( strcmp( pszCFGLeft, "NAC" ) == 0 ) + strcpy( szaHeadLists, "0NN 0NNS 0NNP 0NNPS 0NP 0NAC 0EX 0$ 0CD 0QP + 0PRP 0VBG 0JJ 0JJS 0JJR 0ADJP 0FW 0" ); + else if( strcmp( pszCFGLeft, "PP" ) == 0 ) + strcpy( szaHeadLists, "1IN 1TO 1VBG 1VBN 1RP 1FW 1" ); + else if( strcmp( pszCFGLeft, "PRN" ) == 0 ) + strcpy( szaHeadLists, "1" ); + else if( strcmp( pszCFGLeft, "PRT" ) == 0 ) + strcpy( szaHeadLists, "1RP 1" ); + else if( strcmp( pszCFGLeft, "QP" ) == 0 ) + strcpy( szaHeadLists, "0$ 0IN 0NNS 0NN 0JJ 0RB 0DT 0CD 0NCD 0QP 0JJR + 0JJS 0" ); + else if( strcmp( pszCFGLeft, "RRC" ) == 0 ) + strcpy( szaHeadLists, "1VP 1NP 1ADVP 1ADJP 1PP 1" ); + else if( strcmp( pszCFGLeft, "S" ) == 0 ) + strcpy( szaHeadLists, "0TO 0IN 0VP 0S 0SBAR 0ADJP 0UCP 0NP 0" ); + else if( strcmp( pszCFGLeft, "SBAR" ) == 0 ) + strcpy( szaHeadLists, "0WHNP 0WHPP 0WHADVP 0WHADJP 0IN 0DT 0S 0SQ + 0SINV 0SBAR 0FRAG 0" ); + else if( strcmp( pszCFGLeft, "SBARQ" ) == 0 ) + strcpy( szaHeadLists, "0SQ 0S 0SINV 0SBARQ 0FRAG 0" ); + else if( strcmp( pszCFGLeft, "SINV" ) == 0 ) + strcpy( szaHeadLists, "0VBZ 0VBD 0VBP 0VB 0MD 0VP 0S 0SINV 0ADJP 0NP + 0" ); + else if( strcmp( pszCFGLeft, "SQ" ) == 0 ) + strcpy( szaHeadLists, "0VBZ 0VBD 0VBP 0VB 0MD 0VP 0SQ 0" ); + else if( strcmp( pszCFGLeft, "UCP" ) == 0 ) + strcpy( szaHeadLists, "1" ); + else if( strcmp( pszCFGLeft, "VP" ) == 0 ) + strcpy( szaHeadLists, "0TO 0VBD 0VBN 0MD 0VBZ 0VB 0VBG 0VBP 0VP + 0ADJP 0NN 0NNS 0NP 0" ); + else if( strcmp( pszCFGLeft, "WHADJP" ) == 0 ) + strcpy( szaHeadLists, "0CC 0WRB 0JJ 0ADJP 0" ); + else if( strcmp( pszCFGLeft, "WHADVP" ) == 0 ) + strcpy( szaHeadLists, "1CC 1WRB 1" ); + else if( strcmp( pszCFGLeft, "WHNP" ) == 0 ) + strcpy( szaHeadLists, "0WDT 0WP 0WP$ 0WHADJP 0WHPP 0WHNP 0" ); + else if( strcmp( pszCFGLeft, "WHPP" ) == 0 ) + strcpy( szaHeadLists, "1IN 1TO FW 1" ); + else if( strcmp( pszCFGLeft, "NP" ) == 0 ) + strcpy( szaHeadLists, "0NN NNP NNS NNPS NX POS JJR 0NP 0$ ADJP PRN + 0CD 0JJ JJS RB QP 0" ); + */ + + return fnMyOwnHeadWordRule(szaHeadLists, vectRight); + } + + private: + int fnMyOwnHeadWordRule(char *pszaHeadLists, CVectorStr vectRight) { + char szHeadList[201], *p; + char szTerm[101]; + int J; + + p = pszaHeadLists; + + int iCountRight; + + iCountRight = vectRight.size(); + + szHeadList[0] = '\0'; + while (1) { + szTerm[0] = '\0'; + sscanf(p, "%s", szTerm); + if (strlen(szHeadList) == 0) { + if (strcmp(szTerm, "0") == 0) { + return iCountRight - 1; + } + if (strcmp(szTerm, "1") == 0) { + return 0; + } + + sprintf(szHeadList, "%c %s ", szTerm[0], szTerm + 1); + p = strstr(p, szTerm); + p += strlen(szTerm); + } else { + if ((szTerm[0] == '0') || (szTerm[0] == '1')) { + if (szHeadList[0] == '0') { + for (J = iCountRight - 1; J >= 0; J--) { + sprintf(szTerm, " %s ", vectRight.at(J).c_str()); + if (strstr(szHeadList, szTerm) != NULL) return J; + } + } else { + for (J = 0; J < iCountRight; J++) { + sprintf(szTerm, " %s ", vectRight.at(J).c_str()); + if (strstr(szHeadList, szTerm) != NULL) return J; + } + } + + szHeadList[0] = '\0'; + } else { + strcat(szHeadList, szTerm); + strcat(szHeadList, " "); + + p = strstr(p, szTerm); + p += strlen(szTerm); + } + } + } + + return 0; + } +}; + +struct SParsedTree { + SParsedTree() { m_ptRoot = NULL; } + ~SParsedTree() { + if (m_ptRoot != NULL) delete m_ptRoot; + } + static SParsedTree *fnConvertFromString(const char *pszStr) { + if (strcmp(pszStr, "(())") == 0) return NULL; + SParsedTree *pTree = new SParsedTree(); + + std::vector vecSyn; + fnReadSyntactic(pszStr, vecSyn); + + int iLeft = 1, iRight = 1; //# left/right parenthesis + + STreeItem *pcurrent; + + pTree->m_ptRoot = new STreeItem(vecSyn[1].c_str()); + + pcurrent = pTree->m_ptRoot; + + for (size_t i = 2; i < vecSyn.size() - 1; i++) { + if (strcmp(vecSyn[i].c_str(), "(") == 0) + iLeft++; + else if (strcmp(vecSyn[i].c_str(), ")") == 0) { + iRight++; + if (pcurrent == NULL) { + // error + fprintf(stderr, "ERROR in ConvertFromString\n"); + fprintf(stderr, "%s\n", pszStr); + return NULL; + } + pcurrent = pcurrent->m_ptParent; + } else { + STreeItem *ptNewItem = new STreeItem(vecSyn[i].c_str()); + pcurrent->fnAppend(ptNewItem); + pcurrent = ptNewItem; + + if (strcmp(vecSyn[i - 1].c_str(), "(") != 0 && + strcmp(vecSyn[i - 1].c_str(), ")") != 0) { + pTree->m_vecTerminals.push_back(ptNewItem); + pcurrent = pcurrent->m_ptParent; + } + } + } + + if (iLeft != iRight) { + // error + fprintf(stderr, "the left and right parentheses are not matched!"); + fprintf(stderr, "ERROR in ConvertFromString\n"); + fprintf(stderr, "%s\n", pszStr); + return NULL; + } + + return pTree; + } + + int fnGetNumWord() { return m_vecTerminals.size(); } + + void fnSetSpanInfo() { + int iNextNum = 0; + fnSuffixTraverseSetSpanInfo(m_ptRoot, iNextNum); + } + + void fnSetHeadWord() { + for (size_t i = 0; i < m_vecTerminals.size(); i++) + m_vecTerminals[i]->m_iHeadWord = i; + SGetHeadWord *pGetHeadWord = new SGetHeadWord(); + fnSuffixTraverseSetHeadWord(m_ptRoot, pGetHeadWord); + delete pGetHeadWord; + } + + STreeItem *fnFindNodeForSpan(int iLeft, int iRight, bool bLowest) { + STreeItem *pTreeItem = m_vecTerminals[iLeft]; + + while (pTreeItem->m_iEnd < iRight) { + pTreeItem = pTreeItem->m_ptParent; + if (pTreeItem == NULL) break; + } + if (pTreeItem == NULL) return NULL; + if (pTreeItem->m_iEnd > iRight) return NULL; + + assert(pTreeItem->m_iEnd == iRight); + if (bLowest) return pTreeItem; + + while (pTreeItem->m_ptParent != NULL && + pTreeItem->m_ptParent->fnGetChildrenNum() == 1) + pTreeItem = pTreeItem->m_ptParent; + + return pTreeItem; + } + + private: + void fnSuffixTraverseSetSpanInfo(STreeItem *ptItem, int &iNextNum) { + int I; + int iNumChildren = ptItem->fnGetChildrenNum(); + for (I = 0; I < iNumChildren; I++) + fnSuffixTraverseSetSpanInfo(ptItem->m_vecChildren[I], iNextNum); + + if (I == 0) { + ptItem->m_iBegin = iNextNum; + ptItem->m_iEnd = iNextNum++; + } else { + ptItem->m_iBegin = ptItem->m_vecChildren[0]->m_iBegin; + ptItem->m_iEnd = ptItem->m_vecChildren[I - 1]->m_iEnd; + } + } + + void fnSuffixTraverseSetHeadWord(STreeItem *ptItem, + SGetHeadWord *pGetHeadWord) { + int I, iHeadchild; + + if (ptItem->m_vecChildren.size() == 0) return; + + for (I = 0; I < ptItem->m_vecChildren.size(); I++) + fnSuffixTraverseSetHeadWord(ptItem->m_vecChildren[I], pGetHeadWord); + + std::vector vecRight; + + if (ptItem->m_vecChildren.size() == 1) + iHeadchild = 0; + else { + for (I = 0; I < ptItem->m_vecChildren.size(); I++) + vecRight.push_back(std::string(ptItem->m_vecChildren[I]->m_pszTerm)); + + iHeadchild = pGetHeadWord->fnGetHeadWord(ptItem->m_pszTerm, vecRight); + } + + ptItem->m_iHeadChild = iHeadchild; + ptItem->m_iHeadWord = ptItem->m_vecChildren[iHeadchild]->m_iHeadWord; + } + + static void fnReadSyntactic(const char *pszSyn, + std::vector &vec) { + char *p; + int I; + + int iLeftNum, iRightNum; + char *pszTmp, *pszTerm; + pszTmp = new char[strlen(pszSyn)]; + pszTerm = new char[strlen(pszSyn)]; + pszTmp[0] = pszTerm[0] = '\0'; + + vec.clear(); + + char *pszLine; + pszLine = new char[strlen(pszSyn) + 1]; + strcpy(pszLine, pszSyn); + + char *pszLine2; + + while (1) { + while ((strlen(pszLine) > 0) && (pszLine[strlen(pszLine) - 1] > 0) && + (pszLine[strlen(pszLine) - 1] <= ' ')) + pszLine[strlen(pszLine) - 1] = '\0'; + + if (strlen(pszLine) == 0) break; + + // printf( "%s\n", pszLine ); + pszLine2 = pszLine; + while (pszLine2[0] <= ' ') pszLine2++; + if (pszLine2[0] == '<') continue; + + sscanf(pszLine2 + 1, "%s", pszTmp); + + if (pszLine2[0] == '(') { + iLeftNum = 0; + iRightNum = 0; + } + + p = pszLine2; + while (1) { + pszTerm[0] = '\0'; + sscanf(p, "%s", pszTerm); + + if (strlen(pszTerm) == 0) break; + p = strstr(p, pszTerm); + p += strlen(pszTerm); + + if ((pszTerm[0] == '(') || (pszTerm[strlen(pszTerm) - 1] == ')')) { + if (pszTerm[0] == '(') { + vec.push_back(std::string("(")); + iLeftNum++; + + I = 1; + while (pszTerm[I] == '(' && pszTerm[I] != '\0') { + vec.push_back(std::string("(")); + iLeftNum++; + + I++; + } + + if (strlen(pszTerm) > 1) vec.push_back(std::string(pszTerm + I)); + } else { + char *pTmp; + pTmp = pszTerm + strlen(pszTerm) - 1; + while ((pTmp[0] == ')') && (pTmp >= pszTerm)) pTmp--; + pTmp[1] = '\0'; + + if (strlen(pszTerm) > 0) vec.push_back(std::string(pszTerm)); + pTmp += 2; + + for (I = 0; I <= (int)strlen(pTmp); I++) { + vec.push_back(std::string(")")); + iRightNum++; + } + } + } else { + char *q; + q = strchr(pszTerm, ')'); + if (q != NULL) { + q[0] = '\0'; + if (pszTerm[0] != '\0') vec.push_back(std::string(pszTerm)); + vec.push_back(std::string(")")); + iRightNum++; + + q++; + while (q[0] == ')') { + vec.push_back(std::string(")")); + q++; + iRightNum++; + } + + while (q[0] == '(') { + vec.push_back(std::string("(")); + q++; + iLeftNum++; + } + + if (q[0] != '\0') vec.push_back(std::string(q)); + } else + vec.push_back(std::string(pszTerm)); + } + } + + if (iLeftNum != iRightNum) { + fprintf(stderr, "%s\n", pszSyn); + assert(iLeftNum == iRightNum); + } + /*if ( iLeftNum != iRightNum ) { + printf( "ERROR: left( and right ) is not matched, %d ( and %d + )\n", iLeftNum, iRightNum ); + return; + }*/ + + if (vec.size() >= 2 && strcmp(vec[1].c_str(), "(") == 0) { + //( (IP..) ) + std::vector::iterator it; + it = vec.begin(); + it++; + vec.insert(it, std::string("ROOT")); + } + + break; + } + + delete[] pszLine; + delete[] pszTmp; + delete[] pszTerm; + } + + public: + STreeItem *m_ptRoot; + std::vector m_vecTerminals; // the leaf nodes +}; + +struct SParseReader { + SParseReader(const char *pszParse_Fname, bool bFlattened = false) + : m_bFlattened(bFlattened) { + m_fpIn = fopen(pszParse_Fname, "r"); + assert(m_fpIn != NULL); + } + ~SParseReader() { + if (m_fpIn != NULL) fclose(m_fpIn); + } + + SParsedTree *fnReadNextParseTree() { + SParsedTree *pTree = NULL; + char *pszLine = new char[100001]; + int iLen; + + while (fnReadNextSentence(pszLine, &iLen) == true) { + if (iLen == 0) continue; + + pTree = SParsedTree::fnConvertFromString(pszLine); + if (pTree == NULL) break; + if (m_bFlattened) + fnPostProcessingFlattenedParse(pTree); + else { + pTree->fnSetSpanInfo(); + pTree->fnSetHeadWord(); + } + break; + } + + delete[] pszLine; + return pTree; + } + + SParsedTree *fnReadNextParseTreeWithProb(double *pProb) { + SParsedTree *pTree = NULL; + char *pszLine = new char[100001]; + int iLen; + + while (fnReadNextSentence(pszLine, &iLen) == true) { + if (iLen == 0) continue; + + char *p = strchr(pszLine, ' '); + assert(p != NULL); + p[0] = '\0'; + p++; + if (pProb) (*pProb) = atof(pszLine); + + pTree = SParsedTree::fnConvertFromString(p); + if (m_bFlattened) + fnPostProcessingFlattenedParse(pTree); + else { + pTree->fnSetSpanInfo(); + pTree->fnSetHeadWord(); + } + break; + } + + delete[] pszLine; + return pTree; + } + + private: + /* + * since to the parse tree is a flattened tree, use the head mark to identify + * head info. + * the head node will be marked as "*XP*" + */ + void fnSetParseTreeHeadInfo(SParsedTree *pTree) { + for (size_t i = 0; i < pTree->m_vecTerminals.size(); i++) + pTree->m_vecTerminals[i]->m_iHeadWord = i; + fnSuffixTraverseSetHeadWord(pTree->m_ptRoot); + } + + void fnSuffixTraverseSetHeadWord(STreeItem *pTreeItem) { + if (pTreeItem->m_vecChildren.size() == 0) return; + + for (size_t i = 0; i < pTreeItem->m_vecChildren.size(); i++) + fnSuffixTraverseSetHeadWord(pTreeItem->m_vecChildren[i]); + + std::vector vecRight; + + int iHeadchild; + + if (pTreeItem->fnIsPreTerminal()) { + iHeadchild = 0; + } else { + size_t i; + for (i = 0; i < pTreeItem->m_vecChildren.size(); i++) { + char *p = pTreeItem->m_vecChildren[i]->m_pszTerm; + if (p[0] == '*' && p[strlen(p) - 1] == '*') { + iHeadchild = i; + p[strlen(p) - 1] = '\0'; + std::string str = p + 1; + strcpy(p, str.c_str()); // erase the "*..*" + break; + } + } + assert(i < pTreeItem->m_vecChildren.size()); + } + + pTreeItem->m_iHeadChild = iHeadchild; + pTreeItem->m_iHeadWord = pTreeItem->m_vecChildren[iHeadchild]->m_iHeadWord; + } + void fnPostProcessingFlattenedParse(SParsedTree *pTree) { + pTree->fnSetSpanInfo(); + fnSetParseTreeHeadInfo(pTree); + } + bool fnReadNextSentence(char *pszLine, int *piLength) { + if (feof(m_fpIn) == true) return false; + + int iLen; + + pszLine[0] = '\0'; + + fgets(pszLine, 10001, m_fpIn); + iLen = strlen(pszLine); + while (iLen > 0 && pszLine[iLen - 1] > 0 && pszLine[iLen - 1] < 33) { + pszLine[iLen - 1] = '\0'; + iLen--; + } + + if (piLength != NULL) (*piLength) = iLen; + + return true; + } + + private: + FILE *m_fpIn; + const bool m_bFlattened; +}; + +/* + * Note: + * m_vec_s_align.size() may not be equal to the length of source side + *sentence + * due to the last words may not be aligned + * + */ +struct SAlignment { + typedef std::vector SingleAlign; + SAlignment(const char* pszAlign) { fnInitializeAlignment(pszAlign); } + ~SAlignment() {} + + bool fnIsAligned(int i, bool s) const { + const std::vector* palign; + if (s == true) + palign = &m_vec_s_align; + else + palign = &m_vec_t_align; + if ((*palign)[i].size() == 0) return false; + return true; + } + + /* + * return true if [b, e] is aligned phrases on source side (if s==true) or on + * the target side (if s==false); + * return false, otherwise. + */ + bool fnIsAlignedPhrase(int b, int e, bool s, int* pob, int* poe) const { + int ob, oe; //[b, e] on the other side + if (s == true) + fnGetLeftRightMost(b, e, m_vec_s_align, ob, oe); + else + fnGetLeftRightMost(b, e, m_vec_t_align, ob, oe); + + if (ob == -1) { + if (pob != NULL) (*pob) = -1; + if (poe != NULL) (*poe) = -1; + return false; // no aligned word among [b, e] + } + if (pob != NULL) (*pob) = ob; + if (poe != NULL) (*poe) = oe; + + int bb, be; //[b, e] back given [ob, oe] on the other side + if (s == true) + fnGetLeftRightMost(ob, oe, m_vec_t_align, bb, be); + else + fnGetLeftRightMost(ob, oe, m_vec_s_align, bb, be); + + if (bb < b || be > e) return false; + return true; + } + + bool fnIsAlignedTightPhrase(int b, int e, bool s, int* pob, int* poe) const { + const std::vector* palign; + if (s == true) + palign = &m_vec_s_align; + else + palign = &m_vec_t_align; + + if ((*palign).size() <= e || (*palign)[b].size() == 0 || + (*palign)[e].size() == 0) + return false; + + return fnIsAlignedPhrase(b, e, s, pob, poe); + } + + void fnGetLeftRightMost(int b, int e, bool s, int& ob, int& oe) const { + if (s == true) + fnGetLeftRightMost(b, e, m_vec_s_align, ob, oe); + else + fnGetLeftRightMost(b, e, m_vec_t_align, ob, oe); + } + + /* + * look the translation of source[b, e] is continuous or not + * 1) return "Unaligned": if the source[b, e] is translated silently; + * 2) return "Con't": if none of target words in target[.., ..] is exclusively + * aligned to any word outside source[b, e] + * 3) return "Discon't": otherwise; + */ + std::string fnIsContinuous(int b, int e) const { + int ob, oe; + fnGetLeftRightMost(b, e, true, ob, oe); + if (ob == -1) return "Unaligned"; + + for (int i = ob; i <= oe; i++) { + if (!fnIsAligned(i, false)) continue; + const SingleAlign& a = m_vec_t_align[i]; + int j; + for (j = 0; j < a.size(); j++) + if (a[j] >= b && a[j] <= e) break; + if (j == a.size()) return "Discon't"; + } + return "Con't"; + } + + const SingleAlign* fnGetSingleWordAlign(int i, bool s) const { + if (s == true) { + if (i >= m_vec_s_align.size()) return NULL; + return &(m_vec_s_align[i]); + } else { + if (i >= m_vec_t_align.size()) return NULL; + return &(m_vec_t_align[i]); + } + } + + private: + void fnGetLeftRightMost(int b, int e, const std::vector& align, + int& ob, int& oe) const { + ob = oe = -1; + for (int i = b; i <= e && i < align.size(); i++) { + if (align[i].size() > 0) { + if (align[i][0] < ob || ob == -1) ob = align[i][0]; + if (oe < align[i][align[i].size() - 1]) + oe = align[i][align[i].size() - 1]; + } + } + } + void fnInitializeAlignment(const char* pszAlign) { + m_vec_s_align.clear(); + m_vec_t_align.clear(); + + std::vector terms = SplitOnWhitespace(std::string(pszAlign)); + int si, ti; + for (size_t i = 0; i < terms.size(); i++) { + sscanf(terms[i].c_str(), "%d-%d", &si, &ti); + + while (m_vec_s_align.size() <= si) { + SingleAlign sa; + m_vec_s_align.push_back(sa); + } + while (m_vec_t_align.size() <= ti) { + SingleAlign sa; + m_vec_t_align.push_back(sa); + } + + m_vec_s_align[si].push_back(ti); + m_vec_t_align[ti].push_back(si); + } + + // sort + for (size_t i = 0; i < m_vec_s_align.size(); i++) { + std::sort(m_vec_s_align[i].begin(), m_vec_s_align[i].end()); + } + for (size_t i = 0; i < m_vec_t_align.size(); i++) { + std::sort(m_vec_t_align[i].begin(), m_vec_t_align[i].end()); + } + } + + private: + std::vector m_vec_s_align; // source side words' alignment + std::vector m_vec_t_align; // target side words' alignment +}; + +struct SAlignmentReader { + SAlignmentReader(const char* pszFname) { + m_fpIn = fopen(pszFname, "r"); + assert(m_fpIn != NULL); + } + ~SAlignmentReader() { + if (m_fpIn != NULL) fclose(m_fpIn); + } + SAlignment* fnReadNextAlignment() { + if (feof(m_fpIn) == true) return NULL; + char* pszLine = new char[100001]; + pszLine[0] = '\0'; + fgets(pszLine, 10001, m_fpIn); + int iLen = strlen(pszLine); + if (iLen == 0) return NULL; + while (iLen > 0 && pszLine[iLen - 1] > 0 && pszLine[iLen - 1] < 33) { + pszLine[iLen - 1] = '\0'; + iLen--; + } + SAlignment* pAlign = new SAlignment(pszLine); + delete[] pszLine; + return pAlign; + } + + private: + FILE* m_fpIn; +}; + +struct SArgument { + SArgument(const char* pszRole, int iBegin, int iEnd, float fProb) { + m_pszRole = new char[strlen(pszRole) + 1]; + strcpy(m_pszRole, pszRole); + m_iBegin = iBegin; + m_iEnd = iEnd; + m_fProb = fProb; + m_pTreeItem = NULL; + } + ~SArgument() { delete[] m_pszRole; } + + void fnSetTreeItem(STreeItem* pTreeItem) { + m_pTreeItem = pTreeItem; + if (m_pTreeItem != NULL && m_pTreeItem->m_iBegin != -1) { + assert(m_pTreeItem->m_iBegin == m_iBegin); + assert(m_pTreeItem->m_iEnd == m_iEnd); + } + } + + char* m_pszRole; // argument rule, e.g., ARG0, ARGM-TMP + int m_iBegin; + int m_iEnd; // the span of the argument, [m_iBegin, m_iEnd] + float m_fProb; // the probability of this role, + STreeItem* m_pTreeItem; +}; + +struct SPredicate { + SPredicate(const char* pszLemma, int iPosition) { + if (pszLemma != NULL) { + m_pszLemma = new char[strlen(pszLemma) + 1]; + strcpy(m_pszLemma, pszLemma); + } else + m_pszLemma = NULL; + m_iPosition = iPosition; + } + ~SPredicate() { + if (m_pszLemma != NULL) delete[] m_pszLemma; + for (size_t i = 0; i < m_vecArgt.size(); i++) delete m_vecArgt[i]; + } + int fnAppend(const char* pszRole, int iBegin, int iEnd) { + SArgument* pArgt = new SArgument(pszRole, iBegin, iEnd, 1.0); + return fnAppend(pArgt); + } + int fnAppend(SArgument* pArgt) { + m_vecArgt.push_back(pArgt); + int iPosition = m_vecArgt.size() - 1; + return iPosition; + } + + char* m_pszLemma; // lemma of the predicate, for Chinese, it's always as same + // as the predicate itself + int m_iPosition; // the position in sentence + std::vector m_vecArgt; // arguments associated to the predicate +}; + +struct SSrlSentence { + SSrlSentence() { m_pTree = NULL; } + ~SSrlSentence() { + if (m_pTree != NULL) delete m_pTree; + + for (size_t i = 0; i < m_vecPred.size(); i++) delete m_vecPred[i]; + } + int fnAppend(const char* pszLemma, int iPosition) { + SPredicate* pPred = new SPredicate(pszLemma, iPosition); + return fnAppend(pPred); + } + int fnAppend(SPredicate* pPred) { + m_vecPred.push_back(pPred); + int iPosition = m_vecPred.size() - 1; + return iPosition; + } + int GetPredicateNum() { return m_vecPred.size(); } + + SParsedTree* m_pTree; + std::vector m_vecPred; +}; + +struct SSrlSentenceReader { + SSrlSentenceReader(const char* pszSrlFname) { + m_fpIn = fopen(pszSrlFname, "r"); + assert(m_fpIn != NULL); + } + ~SSrlSentenceReader() { + if (m_fpIn != NULL) fclose(m_fpIn); + } + + inline void fnReplaceAll(std::string& str, const std::string& from, + const std::string& to) { + size_t start_pos = 0; + while ((start_pos = str.find(from, start_pos)) != std::string::npos) { + str.replace(start_pos, from.length(), to); + start_pos += to.length(); // In case 'to' contains 'from', like replacing + // 'x' with 'yx' + } + } + + // TODO: here only considers flat predicate-argument structure + // i.e., no overlap among them + SSrlSentence* fnReadNextSrlSentence() { + std::vector > vecContent; + if (fnReadNextContent(vecContent) == false) return NULL; + + SSrlSentence* pSrlSentence = new SSrlSentence(); + int iSize = vecContent.size(); + // put together syntactic text + std::ostringstream ostr; + for (int i = 0; i < iSize; i++) { + std::string strSynSeg = + vecContent[i][5]; // the 5th column is the syntactic segment + size_t iPosition = strSynSeg.find_first_of('*'); + assert(iPosition != std::string::npos); + std::ostringstream ostrTmp; + ostrTmp << "(" << vecContent[i][2] << " " << vecContent[i][0] + << ")"; // the 2th column is POS-tag, and the 0th column is word + strSynSeg.replace(iPosition, 1, ostrTmp.str()); + fnReplaceAll(strSynSeg, "(", " ("); + ostr << strSynSeg; + } + std::string strSyn = ostr.str(); + pSrlSentence->m_pTree = SParsedTree::fnConvertFromString(strSyn.c_str()); + pSrlSentence->m_pTree->fnSetHeadWord(); + pSrlSentence->m_pTree->fnSetSpanInfo(); + + // read predicate-argument structure + int iNumPred = vecContent[0].size() - 8; + for (int i = 0; i < iNumPred; i++) { + std::vector vecRole; + std::vector vecBegin; + std::vector vecEnd; + int iPred = -1; + for (int j = 0; j < iSize; j++) { + const char* p = vecContent[j][i + 8].c_str(); + const char* q; + if (p[0] == '(') { + // starting position of an argument(or predicate) + vecBegin.push_back(j); + q = strchr(p, '*'); + assert(q != NULL); + vecRole.push_back(vecContent[j][i + 8].substr(1, q - p - 1)); + if (vecRole.back().compare("V") == 0) { + assert(iPred == -1); + iPred = vecRole.size() - 1; + } + } + if (p[strlen(p) - 1] == ')') { + // end position of an argument(or predicate) + vecEnd.push_back(j); + assert(vecBegin.size() == vecEnd.size()); + } + } + assert(iPred != -1); + SPredicate* pPred = new SPredicate( + pSrlSentence->m_pTree->m_vecTerminals[vecBegin[iPred]]->m_pszTerm, + vecBegin[iPred]); + pSrlSentence->fnAppend(pPred); + for (size_t j = 0; j < vecBegin.size(); j++) { + if (j == iPred) continue; + pPred->fnAppend(vecRole[j].c_str(), vecBegin[j], vecEnd[j]); + pPred->m_vecArgt.back()->fnSetTreeItem( + pSrlSentence->m_pTree->fnFindNodeForSpan(vecBegin[j], vecEnd[j], + false)); + } + } + return pSrlSentence; + } + + private: + bool fnReadNextContent(std::vector >& vecContent) { + vecContent.clear(); + if (feof(m_fpIn) == true) return false; + char* pszLine; + pszLine = new char[100001]; + pszLine[0] = '\0'; + int iLen; + while (!feof(m_fpIn)) { + fgets(pszLine, 10001, m_fpIn); + iLen = strlen(pszLine); + while (iLen > 0 && pszLine[iLen - 1] > 0 && pszLine[iLen - 1] < 33) { + pszLine[iLen - 1] = '\0'; + iLen--; + } + if (iLen == 0) break; // end of this sentence + + std::vector terms = SplitOnWhitespace(std::string(pszLine)); + assert(terms.size() > 7); + vecContent.push_back(terms); + } + delete[] pszLine; + return true; + } + + private: + FILE* m_fpIn; +}; + +typedef std::unordered_map Map; +typedef std::unordered_map::iterator Iterator; + +struct Tsuruoka_Maxent { + Tsuruoka_Maxent(const char* pszModelFName) { + if (pszModelFName != NULL) { + m_pModel = new ME_Model(); + m_pModel->load_from_file(pszModelFName); + } else + m_pModel = NULL; + } + + ~Tsuruoka_Maxent() { + if (m_pModel != NULL) delete m_pModel; + } + + void fnTrain(const char* pszInstanceFName, const char* pszAlgorithm, + const char* pszModelFName, int /*iNumIteration*/) { + assert(strcmp(pszAlgorithm, "l1") == 0 || strcmp(pszAlgorithm, "l2") == 0 || + strcmp(pszAlgorithm, "sgd") == 0 || + strcmp(pszAlgorithm, "SGD") == 0); + FILE* fpIn = fopen(pszInstanceFName, "r"); + + ME_Model* pModel = new ME_Model(); + + char* pszLine = new char[100001]; + int iNumInstances = 0; + int iLen; + while (!feof(fpIn)) { + pszLine[0] = '\0'; + fgets(pszLine, 20000, fpIn); + if (strlen(pszLine) == 0) { + continue; + } + + iLen = strlen(pszLine); + while (iLen > 0 && pszLine[iLen - 1] > 0 && pszLine[iLen - 1] < 33) { + pszLine[iLen - 1] = '\0'; + iLen--; + } + + iNumInstances++; + + ME_Sample* pmes = new ME_Sample(); + + char* p = strrchr(pszLine, ' '); + assert(p != NULL); + p[0] = '\0'; + p++; + std::vector vecContext; + SplitOnWhitespace(std::string(pszLine), &vecContext); + + pmes->label = std::string(p); + for (size_t i = 0; i < vecContext.size(); i++) + pmes->add_feature(vecContext[i]); + pModel->add_training_sample((*pmes)); + if (iNumInstances % 100000 == 0) + fprintf(stdout, "......Reading #Instances: %1d\n", iNumInstances); + delete pmes; + } + fprintf(stdout, "......Reading #Instances: %1d\n", iNumInstances); + fclose(fpIn); + + if (strcmp(pszAlgorithm, "l1") == 0) + pModel->use_l1_regularizer(1.0); + else if (strcmp(pszAlgorithm, "l2") == 0) + pModel->use_l2_regularizer(1.0); + else + pModel->use_SGD(); + + pModel->train(); + pModel->save_to_file(pszModelFName); + + delete pModel; + fprintf(stdout, "......Finished Training\n"); + fprintf(stdout, "......Model saved as %s\n", pszModelFName); + delete[] pszLine; + } + + double fnEval(const char* pszContext, const char* pszOutcome) const { + std::vector vecContext; + ME_Sample* pmes = new ME_Sample(); + SplitOnWhitespace(std::string(pszContext), &vecContext); + + for (size_t i = 0; i < vecContext.size(); i++) + pmes->add_feature(vecContext[i]); + std::vector vecProb = m_pModel->classify(*pmes); + delete pmes; + int iLableID = m_pModel->get_class_id(pszOutcome); + return vecProb[iLableID]; + } + void fnEval(const char* pszContext, + std::vector >& vecOutput) const { + std::vector vecContext; + ME_Sample* pmes = new ME_Sample(); + SplitOnWhitespace(std::string(pszContext), &vecContext); + + vecOutput.clear(); + + for (size_t i = 0; i < vecContext.size(); i++) + pmes->add_feature(vecContext[i]); + std::vector vecProb = m_pModel->classify(*pmes); + + for (size_t i = 0; i < vecProb.size(); i++) { + std::string label = m_pModel->get_class_label(i); + vecOutput.push_back(make_pair(label, vecProb[i])); + } + delete pmes; + } + void fnEval(const char* pszContext, std::vector& vecOutput) const { + std::vector vecContext; + ME_Sample* pmes = new ME_Sample(); + SplitOnWhitespace(std::string(pszContext), &vecContext); + + vecOutput.clear(); + + for (size_t i = 0; i < vecContext.size(); i++) + pmes->add_feature(vecContext[i]); + std::vector vecProb = m_pModel->classify(*pmes); + + for (size_t i = 0; i < vecProb.size(); i++) { + std::string label = m_pModel->get_class_label(i); + vecOutput.push_back(vecProb[i]); + } + delete pmes; + } + int fnGetClassId(const std::string& strLabel) const { + return m_pModel->get_class_id(strLabel); + } + + private: + ME_Model* m_pModel; +}; + +// an argument item or a predicate item (the verb itself) +struct SSRLItem { + SSRLItem(const STreeItem *tree_item, std::string role) + : tree_item_(tree_item), role_(role) {} + ~SSRLItem() {} + const STreeItem *tree_item_; + const std::string role_; +}; + +struct SPredicateItem { + SPredicateItem(const SParsedTree *tree, const SPredicate *pred) + : pred_(pred) { + vec_items_.reserve(pred->m_vecArgt.size() + 1); + for (int i = 0; i < pred->m_vecArgt.size(); i++) { + vec_items_.push_back( + new SSRLItem(pred->m_vecArgt[i]->m_pTreeItem, + std::string(pred->m_vecArgt[i]->m_pszRole))); + } + vec_items_.push_back( + new SSRLItem(tree->m_vecTerminals[pred->m_iPosition]->m_ptParent, + std::string("Pred"))); + sort(vec_items_.begin(), vec_items_.end(), SortFunction); + + begin_ = vec_items_[0]->tree_item_->m_iBegin; + end_ = vec_items_[vec_items_.size() - 1]->tree_item_->m_iEnd; + } + + ~SPredicateItem() { vec_items_.clear(); } + + static bool SortFunction(SSRLItem *i, SSRLItem *j) { + return (i->tree_item_->m_iBegin < j->tree_item_->m_iBegin); + } + + std::vector vec_items_; + int begin_; + int end_; + const SPredicate *pred_; +}; + +struct SArgumentReorderModel { + public: + static std::string fnGetBlockOutcome(int iBegin, int iEnd, + SAlignment *pAlign) { + return pAlign->fnIsContinuous(iBegin, iEnd); + } + static void fnGetReorderType(SPredicateItem *pPredItem, SAlignment *pAlign, + std::vector &vecStrLeftReorder, + std::vector &vecStrRightReorder) { + std::vector vecLeft, vecRight; + for (int i = 0; i < pPredItem->vec_items_.size(); i++) { + const STreeItem *pCon1 = pPredItem->vec_items_[i]->tree_item_; + int iLeft1, iRight1; + pAlign->fnGetLeftRightMost(pCon1->m_iBegin, pCon1->m_iEnd, true, iLeft1, + iRight1); + vecLeft.push_back(iLeft1); + vecRight.push_back(iRight1); + } + std::vector vecLeftPosition; + fnGetRelativePosition(vecLeft, vecLeftPosition); + std::vector vecRightPosition; + fnGetRelativePosition(vecRight, vecRightPosition); + + vecStrLeftReorder.clear(); + vecStrRightReorder.clear(); + for (int i = 1; i < vecLeftPosition.size(); i++) { + std::string strOutcome; + fnGetOutcome(vecLeftPosition[i - 1], vecLeftPosition[i], strOutcome); + vecStrLeftReorder.push_back(strOutcome); + fnGetOutcome(vecRightPosition[i - 1], vecRightPosition[i], strOutcome); + vecStrRightReorder.push_back(strOutcome); + } + } + + /* + * features: + * f1: (left_label, right_label, parent_label) + * f2: (left_label, right_label, parent_label, other_right_sibling_label) + * f3: (left_label, right_label, parent_label, other_left_sibling_label) + * f4: (left_label, right_label, left_head_pos) + * f5: (left_label, right_label, left_head_word) + * f6: (left_label, right_label, right_head_pos) + * f7: (left_label, right_label, right_head_word) + * f8: (left_label, right_label, left_chunk_status) + * f9: (left_label, right_label, right_chunk_status) + * f10: (left_label, parent_label) + * f11: (right_label, parent_label) + * + * f1: (left_role, right_role, predicate_term) + * f2: (left_role, right_role, predicate_term, other_right_role) + * f3: (left_role, right_role, predicate_term, other_left_role) + * f4: (left_role, right_role, left_head_pos) + * f5: (left_role, right_role, left_head_word) + * f6: (left_role, right_role, left_syntactic_label) + * f7: (left_role, right_role, right_head_pos) + * f8: (left_role, right_role, right_head_word) + * f8: (left_role, right_role, right_syntactic_label) + * f8: (left_role, right_role, left_chunk_status) + * f9: (left_role, right_role, right_chunk_status) + * f10: (left_role, right_role, left_chunk_status) + * f11: (left_role, right_role, right_chunk_status) + * f12: (left_label, parent_label) + * f13: (right_label, parent_label) + */ + static void fnGenerateFeature(const SParsedTree *pTree, + const SPredicate *pPred, + const SPredicateItem *pPredItem, int iPos, + const std::string &strBlock1, + const std::string &strBlock2, + std::ostringstream &ostr) { + SSRLItem *pSRLItem1 = pPredItem->vec_items_[iPos - 1]; + SSRLItem *pSRLItem2 = pPredItem->vec_items_[iPos]; + const STreeItem *pCon1 = pSRLItem1->tree_item_; + const STreeItem *pCon2 = pSRLItem2->tree_item_; + + std::string left_role = pSRLItem1->role_; + std::string right_role = pSRLItem2->role_; + + std::string predicate_term = + pTree->m_vecTerminals[pPred->m_iPosition]->m_pszTerm; + + std::vector vec_other_right_sibling; + for (int i = iPos + 1; i < pPredItem->vec_items_.size(); i++) + vec_other_right_sibling.push_back( + std::string(pPredItem->vec_items_[i]->role_)); + if (vec_other_right_sibling.size() == 0) + vec_other_right_sibling.push_back(std::string("NULL")); + + std::vector vec_other_left_sibling; + for (int i = 0; i < iPos - 1; i++) + vec_other_right_sibling.push_back( + std::string(pPredItem->vec_items_[i]->role_)); + if (vec_other_left_sibling.size() == 0) + vec_other_left_sibling.push_back(std::string("NULL")); + + // generate features + // f1 + ostr << "f1=" << left_role << "_" << right_role << "_" << predicate_term; + ostr << "f1=" << left_role << "_" << right_role; + + // f2 + for (int i = 0; i < vec_other_right_sibling.size(); i++) { + ostr << " f2=" << left_role << "_" << right_role << "_" << predicate_term + << "_" << vec_other_right_sibling[i]; + ostr << " f2=" << left_role << "_" << right_role << "_" + << vec_other_right_sibling[i]; + } + // f3 + for (int i = 0; i < vec_other_left_sibling.size(); i++) { + ostr << " f3=" << left_role << "_" << right_role << "_" << predicate_term + << "_" << vec_other_left_sibling[i]; + ostr << " f3=" << left_role << "_" << right_role << "_" + << vec_other_left_sibling[i]; + } + // f4 + ostr << " f4=" << left_role << "_" << right_role << "_" + << pTree->m_vecTerminals[pCon1->m_iHeadWord]->m_ptParent->m_pszTerm; + // f5 + ostr << " f5=" << left_role << "_" << right_role << "_" + << pTree->m_vecTerminals[pCon1->m_iHeadWord]->m_pszTerm; + // f6 + ostr << " f6=" << left_role << "_" << right_role << "_" << pCon2->m_pszTerm; + // f7 + ostr << " f7=" << left_role << "_" << right_role << "_" + << pTree->m_vecTerminals[pCon2->m_iHeadWord]->m_ptParent->m_pszTerm; + // f8 + ostr << " f8=" << left_role << "_" << right_role << "_" + << pTree->m_vecTerminals[pCon2->m_iHeadWord]->m_pszTerm; + // f9 + ostr << " f9=" << left_role << "_" << right_role << "_" << pCon2->m_pszTerm; + // f10 + ostr << " f10=" << left_role << "_" << right_role << "_" << strBlock1; + // f11 + ostr << " f11=" << left_role << "_" << right_role << "_" << strBlock2; + // f12 + ostr << " f12=" << left_role << "_" << predicate_term; + ostr << " f12=" << left_role; + // f13 + ostr << " f13=" << right_role << "_" << predicate_term; + ostr << " f13=" << right_role; + } + + private: + static void fnGetOutcome(int i1, int i2, std::string &strOutcome) { + assert(i1 != i2); + if (i1 < i2) { + if (i2 > i1 + 1) + strOutcome = std::string("DM"); + else + strOutcome = std::string("M"); + } else { + if (i1 > i2 + 1) + strOutcome = std::string("DS"); + else + strOutcome = std::string("S"); + } + } + + static void fnGetRelativePosition(const std::vector &vecLeft, + std::vector &vecPosition) { + vecPosition.clear(); + + std::vector vec; + for (int i = 0; i < vecLeft.size(); i++) { + if (vecLeft[i] == -1) { + if (i == 0) + vec.push_back(-1); + else + vec.push_back(vecLeft[i - 1] + 0.1); + } else + vec.push_back(vecLeft[i]); + } + + for (int i = 0; i < vecLeft.size(); i++) { + int count = 0; + + for (int j = 0; j < vecLeft.size(); j++) { + if (j == i) continue; + if (vec[j] < vec[i]) { + count++; + } else if (vec[j] == vec[i] && j < i) { + count++; + } + } + vecPosition.push_back(count); + } + } +}; +} // namespace const_reorder + +#endif // _FF_CONST_REORDER_COMMON_H -- cgit v1.2.3 From 62249e8de1be27057649aa787b715af5727f8a7c Mon Sep 17 00:00:00 2001 From: "Wu, Ke" Date: Wed, 17 Dec 2014 15:41:32 -0500 Subject: Move training routine out of ff_const_reorder_common.h --- decoder/ff_const_reorder_common.h | 93 ---------------------- training/const_reorder/Makefile.am | 8 +- training/const_reorder/argument_reorder_model.cc | 6 +- .../const_reorder/constituent_reorder_model.cc | 6 +- training/const_reorder/trainer.cc | 67 ++++++++++++++++ training/const_reorder/trainer.h | 12 +++ 6 files changed, 91 insertions(+), 101 deletions(-) create mode 100644 training/const_reorder/trainer.cc create mode 100644 training/const_reorder/trainer.h (limited to 'decoder/ff_const_reorder_common.h') diff --git a/decoder/ff_const_reorder_common.h b/decoder/ff_const_reorder_common.h index 7c111de3..b124ce47 100644 --- a/decoder/ff_const_reorder_common.h +++ b/decoder/ff_const_reorder_common.h @@ -1091,99 +1091,6 @@ struct Tsuruoka_Maxent { if (m_pModel != NULL) delete m_pModel; } - void fnTrain(const char* pszInstanceFName, const char* pszAlgorithm, - const char* pszModelFName, int /*iNumIteration*/) { - assert(strcmp(pszAlgorithm, "l1") == 0 || strcmp(pszAlgorithm, "l2") == 0 || - strcmp(pszAlgorithm, "sgd") == 0 || - strcmp(pszAlgorithm, "SGD") == 0); - FILE* fpIn = fopen(pszInstanceFName, "r"); - - ME_Model* pModel = new ME_Model(); - - char* pszLine = new char[100001]; - int iNumInstances = 0; - int iLen; - while (!feof(fpIn)) { - pszLine[0] = '\0'; - fgets(pszLine, 20000, fpIn); - if (strlen(pszLine) == 0) { - continue; - } - - iLen = strlen(pszLine); - while (iLen > 0 && pszLine[iLen - 1] > 0 && pszLine[iLen - 1] < 33) { - pszLine[iLen - 1] = '\0'; - iLen--; - } - - iNumInstances++; - - ME_Sample* pmes = new ME_Sample(); - - char* p = strrchr(pszLine, ' '); - assert(p != NULL); - p[0] = '\0'; - p++; - std::vector vecContext; - SplitOnWhitespace(std::string(pszLine), &vecContext); - - pmes->label = std::string(p); - for (size_t i = 0; i < vecContext.size(); i++) - pmes->add_feature(vecContext[i]); - pModel->add_training_sample((*pmes)); - if (iNumInstances % 100000 == 0) - fprintf(stdout, "......Reading #Instances: %1d\n", iNumInstances); - delete pmes; - } - fprintf(stdout, "......Reading #Instances: %1d\n", iNumInstances); - fclose(fpIn); - - if (strcmp(pszAlgorithm, "l1") == 0) - pModel->use_l1_regularizer(1.0); - else if (strcmp(pszAlgorithm, "l2") == 0) - pModel->use_l2_regularizer(1.0); - else - pModel->use_SGD(); - - pModel->train(); - pModel->save_to_file(pszModelFName); - - delete pModel; - fprintf(stdout, "......Finished Training\n"); - fprintf(stdout, "......Model saved as %s\n", pszModelFName); - delete[] pszLine; - } - - double fnEval(const char* pszContext, const char* pszOutcome) const { - std::vector vecContext; - ME_Sample* pmes = new ME_Sample(); - SplitOnWhitespace(std::string(pszContext), &vecContext); - - for (size_t i = 0; i < vecContext.size(); i++) - pmes->add_feature(vecContext[i]); - std::vector vecProb = m_pModel->classify(*pmes); - delete pmes; - int iLableID = m_pModel->get_class_id(pszOutcome); - return vecProb[iLableID]; - } - void fnEval(const char* pszContext, - std::vector >& vecOutput) const { - std::vector vecContext; - ME_Sample* pmes = new ME_Sample(); - SplitOnWhitespace(std::string(pszContext), &vecContext); - - vecOutput.clear(); - - for (size_t i = 0; i < vecContext.size(); i++) - pmes->add_feature(vecContext[i]); - std::vector vecProb = m_pModel->classify(*pmes); - - for (size_t i = 0; i < vecProb.size(); i++) { - std::string label = m_pModel->get_class_label(i); - vecOutput.push_back(make_pair(label, vecProb[i])); - } - delete pmes; - } void fnEval(const char* pszContext, std::vector& vecOutput) const { std::vector vecContext; ME_Sample* pmes = new ME_Sample(); diff --git a/training/const_reorder/Makefile.am b/training/const_reorder/Makefile.am index 2e81e588..367ac904 100644 --- a/training/const_reorder/Makefile.am +++ b/training/const_reorder/Makefile.am @@ -1,8 +1,12 @@ +noinst_LIBRARIES = libtrainer.a + +libtrainer_a_SOURCES = trainer.h trainer.cc + bin_PROGRAMS = const_reorder_model_trainer argument_reorder_model_trainer AM_CPPFLAGS = -I$(top_srcdir) -I$(top_srcdir)/utils -I$(top_srcdir)/decoder const_reorder_model_trainer_SOURCES = constituent_reorder_model.cc -const_reorder_model_trainer_LDADD = ../../utils/libutils.a +const_reorder_model_trainer_LDADD = ../../utils/libutils.a libtrainer.a argument_reorder_model_trainer_SOURCES = argument_reorder_model.cc -argument_reorder_model_trainer_LDADD = ../../utils/libutils.a +argument_reorder_model_trainer_LDADD = ../../utils/libutils.a libtrainer.a diff --git a/training/const_reorder/argument_reorder_model.cc b/training/const_reorder/argument_reorder_model.cc index 54402436..87f2ce2f 100644 --- a/training/const_reorder/argument_reorder_model.cc +++ b/training/const_reorder/argument_reorder_model.cc @@ -14,7 +14,7 @@ #include "utils/filelib.h" -#include "decoder/ff_const_reorder_common.h" +#include "trainer.h" using namespace std; using namespace const_reorder; @@ -93,8 +93,8 @@ struct SArgumentReorderTrainer { strcpy(pszNewInstanceFName, pszInstanceFname); } - Tsuruoka_Maxent* pMaxent = new Tsuruoka_Maxent(NULL); - pMaxent->fnTrain(pszNewInstanceFName, "l1", pszModelFname, 300); + Tsuruoka_Maxent_Trainer* pMaxent = new Tsuruoka_Maxent_Trainer; + pMaxent->fnTrain(pszNewInstanceFName, "l1", pszModelFname); delete pMaxent; if (strcmp(pszNewInstanceFName, pszInstanceFname) != 0) { diff --git a/training/const_reorder/constituent_reorder_model.cc b/training/const_reorder/constituent_reorder_model.cc index 6bec3f0b..d3ad0f2b 100644 --- a/training/const_reorder/constituent_reorder_model.cc +++ b/training/const_reorder/constituent_reorder_model.cc @@ -12,7 +12,7 @@ #include "utils/filelib.h" -#include "decoder/ff_const_reorder_common.h" +#include "trainer.h" using namespace std; using namespace const_reorder; @@ -104,8 +104,8 @@ struct SConstReorderTrainer { pZhangleMaxent->fnTrain(pszInstanceFname, "lbfgs", pszModelFname, 100, 2.0); delete pZhangleMaxent;*/ - Tsuruoka_Maxent* pMaxent = new Tsuruoka_Maxent(NULL); - pMaxent->fnTrain(pszNewInstanceFName, "l1", pszModelFname, 300); + Tsuruoka_Maxent_Trainer* pMaxent = new Tsuruoka_Maxent_Trainer; + pMaxent->fnTrain(pszNewInstanceFName, "l1", pszModelFname); delete pMaxent; if (strcmp(pszNewInstanceFName, pszInstanceFname) != 0) { diff --git a/training/const_reorder/trainer.cc b/training/const_reorder/trainer.cc new file mode 100644 index 00000000..e22a8a66 --- /dev/null +++ b/training/const_reorder/trainer.cc @@ -0,0 +1,67 @@ +#include "trainer.h" + +Tsuruoka_Maxent_Trainer::Tsuruoka_Maxent_Trainer() + : const_reorder::Tsuruoka_Maxent(NULL) {} + +void Tsuruoka_Maxent_Trainer::fnTrain(const char* pszInstanceFName, + const char* pszAlgorithm, + const char* pszModelFName) { + assert(strcmp(pszAlgorithm, "l1") == 0 || strcmp(pszAlgorithm, "l2") == 0 || + strcmp(pszAlgorithm, "sgd") == 0 || strcmp(pszAlgorithm, "SGD") == 0); + FILE* fpIn = fopen(pszInstanceFName, "r"); + + ME_Model* pModel = new ME_Model(); + + char* pszLine = new char[100001]; + int iNumInstances = 0; + int iLen; + while (!feof(fpIn)) { + pszLine[0] = '\0'; + fgets(pszLine, 20000, fpIn); + if (strlen(pszLine) == 0) { + continue; + } + + iLen = strlen(pszLine); + while (iLen > 0 && pszLine[iLen - 1] > 0 && pszLine[iLen - 1] < 33) { + pszLine[iLen - 1] = '\0'; + iLen--; + } + + iNumInstances++; + + ME_Sample* pmes = new ME_Sample(); + + char* p = strrchr(pszLine, ' '); + assert(p != NULL); + p[0] = '\0'; + p++; + std::vector vecContext; + SplitOnWhitespace(std::string(pszLine), &vecContext); + + pmes->label = std::string(p); + for (size_t i = 0; i < vecContext.size(); i++) + pmes->add_feature(vecContext[i]); + pModel->add_training_sample((*pmes)); + if (iNumInstances % 100000 == 0) + fprintf(stdout, "......Reading #Instances: %1d\n", iNumInstances); + delete pmes; + } + fprintf(stdout, "......Reading #Instances: %1d\n", iNumInstances); + fclose(fpIn); + + if (strcmp(pszAlgorithm, "l1") == 0) + pModel->use_l1_regularizer(1.0); + else if (strcmp(pszAlgorithm, "l2") == 0) + pModel->use_l2_regularizer(1.0); + else + pModel->use_SGD(); + + pModel->train(); + pModel->save_to_file(pszModelFName); + + delete pModel; + fprintf(stdout, "......Finished Training\n"); + fprintf(stdout, "......Model saved as %s\n", pszModelFName); + delete[] pszLine; +} diff --git a/training/const_reorder/trainer.h b/training/const_reorder/trainer.h new file mode 100644 index 00000000..e574a536 --- /dev/null +++ b/training/const_reorder/trainer.h @@ -0,0 +1,12 @@ +#ifndef TRAINING_CONST_REORDER_TRAINER_H_ +#define TRAINING_CONST_REORDER_TRAINER_H_ + +#include "decoder/ff_const_reorder_common.h" + +struct Tsuruoka_Maxent_Trainer : const_reorder::Tsuruoka_Maxent { + Tsuruoka_Maxent_Trainer(); + void fnTrain(const char* pszInstanceFName, const char* pszAlgorithm, + const char* pszModelFName); +}; + +#endif // TRAINING_CONST_REORDER_TRAINER_H_ -- cgit v1.2.3 From b6dd5a683db9dda2d634dd2fdb76606819594901 Mon Sep 17 00:00:00 2001 From: "Wu, Ke" Date: Wed, 17 Dec 2014 16:00:04 -0500 Subject: Combine everything related to maxent to a single file --- decoder/ff_const_reorder_common.h | 6 +- training/const_reorder/trainer.cc | 4 +- utils/Makefile.am | 5 - utils/lbfgs.cpp | 108 ---------- utils/lbfgs.h | 20 -- utils/mathvec.h | 87 -------- utils/maxent.cpp | 427 +++++++++++++++++++++++++++++++++++++- utils/maxent.h | 95 ++++++++- utils/owlqn.cpp | 127 ------------ utils/sgd.cpp | 193 ----------------- 10 files changed, 516 insertions(+), 556 deletions(-) delete mode 100644 utils/lbfgs.cpp delete mode 100644 utils/lbfgs.h delete mode 100644 utils/mathvec.h delete mode 100644 utils/owlqn.cpp delete mode 100644 utils/sgd.cpp (limited to 'decoder/ff_const_reorder_common.h') diff --git a/decoder/ff_const_reorder_common.h b/decoder/ff_const_reorder_common.h index b124ce47..755fd948 100644 --- a/decoder/ff_const_reorder_common.h +++ b/decoder/ff_const_reorder_common.h @@ -1081,7 +1081,7 @@ typedef std::unordered_map::iterator Iterator; struct Tsuruoka_Maxent { Tsuruoka_Maxent(const char* pszModelFName) { if (pszModelFName != NULL) { - m_pModel = new ME_Model(); + m_pModel = new maxent::ME_Model(); m_pModel->load_from_file(pszModelFName); } else m_pModel = NULL; @@ -1093,7 +1093,7 @@ struct Tsuruoka_Maxent { void fnEval(const char* pszContext, std::vector& vecOutput) const { std::vector vecContext; - ME_Sample* pmes = new ME_Sample(); + maxent::ME_Sample* pmes = new maxent::ME_Sample(); SplitOnWhitespace(std::string(pszContext), &vecContext); vecOutput.clear(); @@ -1113,7 +1113,7 @@ struct Tsuruoka_Maxent { } private: - ME_Model* m_pModel; + maxent::ME_Model* m_pModel; }; // an argument item or a predicate item (the verb itself) diff --git a/training/const_reorder/trainer.cc b/training/const_reorder/trainer.cc index e22a8a66..89bd7479 100644 --- a/training/const_reorder/trainer.cc +++ b/training/const_reorder/trainer.cc @@ -10,7 +10,7 @@ void Tsuruoka_Maxent_Trainer::fnTrain(const char* pszInstanceFName, strcmp(pszAlgorithm, "sgd") == 0 || strcmp(pszAlgorithm, "SGD") == 0); FILE* fpIn = fopen(pszInstanceFName, "r"); - ME_Model* pModel = new ME_Model(); + maxent::ME_Model* pModel = new maxent::ME_Model(); char* pszLine = new char[100001]; int iNumInstances = 0; @@ -30,7 +30,7 @@ void Tsuruoka_Maxent_Trainer::fnTrain(const char* pszInstanceFName, iNumInstances++; - ME_Sample* pmes = new ME_Sample(); + maxent::ME_Sample* pmes = new maxent::ME_Sample(); char* p = strrchr(pszLine, ' '); assert(p != NULL); diff --git a/utils/Makefile.am b/utils/Makefile.am index fabb4454..e0221e64 100644 --- a/utils/Makefile.am +++ b/utils/Makefile.am @@ -38,11 +38,8 @@ libutils_a_SOURCES = \ have_64_bits.h \ indices_after.h \ kernel_string_subseq.h \ - lbfgs.h \ - lbfgs.cpp \ logval.h \ m.h \ - mathvec.h \ maxent.h \ maxent.cpp \ murmur_hash3.h \ @@ -50,8 +47,6 @@ libutils_a_SOURCES = \ named_enum.h \ null_deleter.h \ null_traits.h \ - owlqn.cpp \ - sgd.cpp \ perfect_hash.h \ prob.h \ sampler.h \ diff --git a/utils/lbfgs.cpp b/utils/lbfgs.cpp deleted file mode 100644 index bd26f048..00000000 --- a/utils/lbfgs.cpp +++ /dev/null @@ -1,108 +0,0 @@ -#include -#include -#include -#include -#include "mathvec.h" -#include "lbfgs.h" -#include "maxent.h" - -using namespace std; - -const static int M = LBFGS_M; -const static double LINE_SEARCH_ALPHA = 0.1; -const static double LINE_SEARCH_BETA = 0.5; - -// stopping criteria -int LBFGS_MAX_ITER = 300; -const static double MIN_GRAD_NORM = 0.0001; - -double ME_Model::backtracking_line_search(const Vec& x0, const Vec& grad0, - const double f0, const Vec& dx, - Vec& x, Vec& grad1) { - double t = 1.0 / LINE_SEARCH_BETA; - - double f; - do { - t *= LINE_SEARCH_BETA; - x = x0 + t * dx; - f = FunctionGradient(x.STLVec(), grad1.STLVec()); - // cout << "*"; - } while (f > f0 + LINE_SEARCH_ALPHA * t * dot_product(dx, grad0)); - - return f; -} - -// -// Jorge Nocedal, "Updating Quasi-Newton Matrices With Limited Storage", -// Mathematics of Computation, Vol. 35, No. 151, pp. 773-782, 1980. -// -Vec approximate_Hg(const int iter, const Vec& grad, const Vec s[], - const Vec y[], const double z[]) { - int offset, bound; - if (iter <= M) { - offset = 0; - bound = iter; - } else { - offset = iter - M; - bound = M; - } - - Vec q = grad; - double alpha[M], beta[M]; - for (int i = bound - 1; i >= 0; i--) { - const int j = (i + offset) % M; - alpha[i] = z[j] * dot_product(s[j], q); - q += -alpha[i] * y[j]; - } - if (iter > 0) { - const int j = (iter - 1) % M; - const double gamma = ((1.0 / z[j]) / dot_product(y[j], y[j])); - // static double gamma; - // if (gamma == 0) gamma = ((1.0 / z[j]) / dot_product(y[j], y[j])); - q *= gamma; - } - for (int i = 0; i <= bound - 1; i++) { - const int j = (i + offset) % M; - beta[i] = z[j] * dot_product(y[j], q); - q += s[j] * (alpha[i] - beta[i]); - } - - return q; -} - -vector ME_Model::perform_LBFGS(const vector& x0) { - const size_t dim = x0.size(); - Vec x = x0; - - Vec grad(dim), dx(dim); - double f = FunctionGradient(x.STLVec(), grad.STLVec()); - - Vec s[M], y[M]; - double z[M]; // rho - - for (int iter = 0; iter < LBFGS_MAX_ITER; iter++) { - - fprintf(stderr, "%3d obj(err) = %f (%6.4f)", iter + 1, -f, _train_error); - if (_nheldout > 0) { - const double heldout_logl = heldout_likelihood(); - fprintf(stderr, " heldout_logl(err) = %f (%6.4f)", heldout_logl, - _heldout_error); - } - fprintf(stderr, "\n"); - - if (sqrt(dot_product(grad, grad)) < MIN_GRAD_NORM) break; - - dx = -1 * approximate_Hg(iter, grad, s, y, z); - - Vec x1(dim), grad1(dim); - f = backtracking_line_search(x, grad, f, dx, x1, grad1); - - s[iter % M] = x1 - x; - y[iter % M] = grad1 - grad; - z[iter % M] = 1.0 / dot_product(y[iter % M], s[iter % M]); - x = x1; - grad = grad1; - } - - return x.STLVec(); -} diff --git a/utils/lbfgs.h b/utils/lbfgs.h deleted file mode 100644 index 4d706f7a..00000000 --- a/utils/lbfgs.h +++ /dev/null @@ -1,20 +0,0 @@ -#ifndef _LBFGS_H_ -#define _LBFGS_H_ - -#include - -// template -// std::vector -// perform_LBFGS(FuncGrad func_grad, const std::vector & x0); - -std::vector perform_LBFGS( - double (*func_grad)(const std::vector &, std::vector &), - const std::vector &x0); - -std::vector perform_OWLQN( - double (*func_grad)(const std::vector &, std::vector &), - const std::vector &x0, const double C); - -const int LBFGS_M = 10; - -#endif diff --git a/utils/mathvec.h b/utils/mathvec.h deleted file mode 100644 index f8c60e5d..00000000 --- a/utils/mathvec.h +++ /dev/null @@ -1,87 +0,0 @@ -#ifndef _MATH_VECTOR_H_ -#define _MATH_VECTOR_H_ - -#include -#include -#include - -class Vec { - private: - std::vector _v; - - public: - Vec(const size_t n = 0, const double val = 0) { _v.resize(n, val); } - Vec(const std::vector& v) : _v(v) {} - const std::vector& STLVec() const { return _v; } - std::vector& STLVec() { return _v; } - size_t Size() const { return _v.size(); } - double& operator[](int i) { return _v[i]; } - const double& operator[](int i) const { return _v[i]; } - Vec& operator+=(const Vec& b) { - assert(b.Size() == _v.size()); - for (size_t i = 0; i < _v.size(); i++) { - _v[i] += b[i]; - } - return *this; - } - Vec& operator*=(const double c) { - for (size_t i = 0; i < _v.size(); i++) { - _v[i] *= c; - } - return *this; - } - void Project(const Vec& y) { - for (size_t i = 0; i < _v.size(); i++) { - // if (sign(_v[i]) != sign(y[i])) _v[i] = 0; - if (_v[i] * y[i] <= 0) _v[i] = 0; - } - } -}; - -inline double dot_product(const Vec& a, const Vec& b) { - double sum = 0; - for (size_t i = 0; i < a.Size(); i++) { - sum += a[i] * b[i]; - } - return sum; -} - -inline std::ostream& operator<<(std::ostream& s, const Vec& a) { - s << "("; - for (size_t i = 0; i < a.Size(); i++) { - if (i != 0) s << ", "; - s << a[i]; - } - s << ")"; - return s; -} - -inline const Vec operator+(const Vec& a, const Vec& b) { - Vec v(a.Size()); - assert(a.Size() == b.Size()); - for (size_t i = 0; i < a.Size(); i++) { - v[i] = a[i] + b[i]; - } - return v; -} - -inline const Vec operator-(const Vec& a, const Vec& b) { - Vec v(a.Size()); - assert(a.Size() == b.Size()); - for (size_t i = 0; i < a.Size(); i++) { - v[i] = a[i] - b[i]; - } - return v; -} - -inline const Vec operator*(const Vec& a, const double c) { - Vec v(a.Size()); - for (size_t i = 0; i < a.Size(); i++) { - v[i] = a[i] * c; - } - return v; -} - -inline const Vec operator*(const double c, const Vec& a) { return a * c; } - -#endif diff --git a/utils/maxent.cpp b/utils/maxent.cpp index 0f49ee9d..fd772e08 100644 --- a/utils/maxent.cpp +++ b/utils/maxent.cpp @@ -3,12 +3,15 @@ */ #include "maxent.h" + +#include +#include #include #include -#include "lbfgs.h" using namespace std; +namespace maxent { double ME_Model::FunctionGradient(const vector& x, vector& grad) { assert((int)_fb.Size() == x.size()); @@ -601,6 +604,428 @@ vector ME_Model::classify(ME_Sample& mes) const { return vp; } +// template +// std::vector +// perform_LBFGS(FuncGrad func_grad, const std::vector & x0); + +std::vector perform_LBFGS( + double (*func_grad)(const std::vector &, std::vector &), + const std::vector &x0); + +std::vector perform_OWLQN( + double (*func_grad)(const std::vector &, std::vector &), + const std::vector &x0, const double C); + +const int LBFGS_M = 10; + +const static int M = LBFGS_M; +const static double LINE_SEARCH_ALPHA = 0.1; +const static double LINE_SEARCH_BETA = 0.5; + +// stopping criteria +int LBFGS_MAX_ITER = 300; +const static double MIN_GRAD_NORM = 0.0001; + +// LBFGS + +double ME_Model::backtracking_line_search(const Vec& x0, const Vec& grad0, + const double f0, const Vec& dx, + Vec& x, Vec& grad1) { + double t = 1.0 / LINE_SEARCH_BETA; + + double f; + do { + t *= LINE_SEARCH_BETA; + x = x0 + t * dx; + f = FunctionGradient(x.STLVec(), grad1.STLVec()); + // cout << "*"; + } while (f > f0 + LINE_SEARCH_ALPHA * t * dot_product(dx, grad0)); + + return f; +} + +// +// Jorge Nocedal, "Updating Quasi-Newton Matrices With Limited Storage", +// Mathematics of Computation, Vol. 35, No. 151, pp. 773-782, 1980. +// +Vec approximate_Hg(const int iter, const Vec& grad, const Vec s[], + const Vec y[], const double z[]) { + int offset, bound; + if (iter <= M) { + offset = 0; + bound = iter; + } else { + offset = iter - M; + bound = M; + } + + Vec q = grad; + double alpha[M], beta[M]; + for (int i = bound - 1; i >= 0; i--) { + const int j = (i + offset) % M; + alpha[i] = z[j] * dot_product(s[j], q); + q += -alpha[i] * y[j]; + } + if (iter > 0) { + const int j = (iter - 1) % M; + const double gamma = ((1.0 / z[j]) / dot_product(y[j], y[j])); + // static double gamma; + // if (gamma == 0) gamma = ((1.0 / z[j]) / dot_product(y[j], y[j])); + q *= gamma; + } + for (int i = 0; i <= bound - 1; i++) { + const int j = (i + offset) % M; + beta[i] = z[j] * dot_product(y[j], q); + q += s[j] * (alpha[i] - beta[i]); + } + + return q; +} + +vector ME_Model::perform_LBFGS(const vector& x0) { + const size_t dim = x0.size(); + Vec x = x0; + + Vec grad(dim), dx(dim); + double f = FunctionGradient(x.STLVec(), grad.STLVec()); + + Vec s[M], y[M]; + double z[M]; // rho + + for (int iter = 0; iter < LBFGS_MAX_ITER; iter++) { + + fprintf(stderr, "%3d obj(err) = %f (%6.4f)", iter + 1, -f, _train_error); + if (_nheldout > 0) { + const double heldout_logl = heldout_likelihood(); + fprintf(stderr, " heldout_logl(err) = %f (%6.4f)", heldout_logl, + _heldout_error); + } + fprintf(stderr, "\n"); + + if (sqrt(dot_product(grad, grad)) < MIN_GRAD_NORM) break; + + dx = -1 * approximate_Hg(iter, grad, s, y, z); + + Vec x1(dim), grad1(dim); + f = backtracking_line_search(x, grad, f, dx, x1, grad1); + + s[iter % M] = x1 - x; + y[iter % M] = grad1 - grad; + z[iter % M] = 1.0 / dot_product(y[iter % M], s[iter % M]); + x = x1; + grad = grad1; + } + + return x.STLVec(); +} + +// OWLQN + +// stopping criteria +int OWLQN_MAX_ITER = 300; + +Vec approximate_Hg(const int iter, const Vec& grad, const Vec s[], + const Vec y[], const double z[]); + +inline int sign(double x) { + if (x > 0) return 1; + if (x < 0) return -1; + return 0; +}; + +static Vec pseudo_gradient(const Vec& x, const Vec& grad0, const double C) { + Vec grad = grad0; + for (size_t i = 0; i < x.Size(); i++) { + if (x[i] != 0) { + grad[i] += C * sign(x[i]); + continue; + } + const double gm = grad0[i] - C; + if (gm > 0) { + grad[i] = gm; + continue; + } + const double gp = grad0[i] + C; + if (gp < 0) { + grad[i] = gp; + continue; + } + grad[i] = 0; + } + + return grad; +} + +double ME_Model::regularized_func_grad(const double C, const Vec& x, + Vec& grad) { + double f = FunctionGradient(x.STLVec(), grad.STLVec()); + for (size_t i = 0; i < x.Size(); i++) { + f += C * fabs(x[i]); + } + + return f; +} + +double ME_Model::constrained_line_search(double C, const Vec& x0, + const Vec& grad0, const double f0, + const Vec& dx, Vec& x, Vec& grad1) { + // compute the orthant to explore + Vec orthant = x0; + for (size_t i = 0; i < orthant.Size(); i++) { + if (orthant[i] == 0) orthant[i] = -grad0[i]; + } + + double t = 1.0 / LINE_SEARCH_BETA; + + double f; + do { + t *= LINE_SEARCH_BETA; + x = x0 + t * dx; + x.Project(orthant); + // for (size_t i = 0; i < x.Size(); i++) { + // if (x0[i] != 0 && sign(x[i]) != sign(x0[i])) x[i] = 0; + // } + + f = regularized_func_grad(C, x, grad1); + // cout << "*"; + } while (f > f0 + LINE_SEARCH_ALPHA * dot_product(x - x0, grad0)); + + return f; +} + +vector ME_Model::perform_OWLQN(const vector& x0, + const double C) { + const size_t dim = x0.size(); + Vec x = x0; + + Vec grad(dim), dx(dim); + double f = regularized_func_grad(C, x, grad); + + Vec s[M], y[M]; + double z[M]; // rho + + for (int iter = 0; iter < OWLQN_MAX_ITER; iter++) { + Vec pg = pseudo_gradient(x, grad, C); + + fprintf(stderr, "%3d obj(err) = %f (%6.4f)", iter + 1, -f, _train_error); + if (_nheldout > 0) { + const double heldout_logl = heldout_likelihood(); + fprintf(stderr, " heldout_logl(err) = %f (%6.4f)", heldout_logl, + _heldout_error); + } + fprintf(stderr, "\n"); + + if (sqrt(dot_product(pg, pg)) < MIN_GRAD_NORM) break; + + dx = -1 * approximate_Hg(iter, pg, s, y, z); + if (dot_product(dx, pg) >= 0) dx.Project(-1 * pg); + + Vec x1(dim), grad1(dim); + f = constrained_line_search(C, x, pg, f, dx, x1, grad1); + + s[iter % M] = x1 - x; + y[iter % M] = grad1 - grad; + z[iter % M] = 1.0 / dot_product(y[iter % M], s[iter % M]); + + x = x1; + grad = grad1; + } + + return x.STLVec(); +} + +// SGD + +// const double SGD_ETA0 = 1; +// const double SGD_ITER = 30; +// const double SGD_ALPHA = 0.85; + +//#define FOLOS_NAIVE +//#define FOLOS_LAZY +#define SGD_CP + +inline void apply_l1_penalty(const int i, const double u, vector& _vl, + vector& q) { + double& w = _vl[i]; + const double z = w; + double& qi = q[i]; + if (w > 0) { + w = max(0.0, w - (u + qi)); + } else if (w < 0) { + w = min(0.0, w + (u - qi)); + } + qi += w - z; +} + +static double l1norm(const vector& v) { + double sum = 0; + for (size_t i = 0; i < v.size(); i++) sum += abs(v[i]); + return sum; +} + +inline void update_folos_lazy(const int iter_sample, const int k, + vector& _vl, + const vector& sum_eta, + vector& last_updated) { + const double penalty = sum_eta[iter_sample] - sum_eta[last_updated[k]]; + double& x = _vl[k]; + if (x > 0) + x = max(0.0, x - penalty); + else + x = min(0.0, x + penalty); + last_updated[k] = iter_sample; +} + +int ME_Model::perform_SGD() { + if (_l2reg > 0) { + cerr << "error: L2 regularization is currently not supported in SGD mode." + << endl; + exit(1); + } + + cerr << "performing SGD" << endl; + + const double l1param = _l1reg; + + const int d = _fb.Size(); + + vector ri(_vs.size()); + for (size_t i = 0; i < ri.size(); i++) ri[i] = i; + + vector grad(d); + int iter_sample = 0; + const double eta0 = SGD_ETA0; + + // cerr << "l1param = " << l1param << endl; + cerr << "eta0 = " << eta0 << " alpha = " << SGD_ALPHA << endl; + + double u = 0; + vector q(d, 0); + vector last_updated(d, 0); + vector sum_eta; + sum_eta.push_back(0); + + for (int iter = 0; iter < SGD_ITER; iter++) { + + random_shuffle(ri.begin(), ri.end()); + + double logl = 0; + int ncorrect = 0, ntotal = 0; + for (size_t i = 0; i < _vs.size(); i++, ntotal++, iter_sample++) { + const Sample& s = _vs[ri[i]]; + +#ifdef FOLOS_LAZY + for (vector::const_iterator j = s.positive_features.begin(); + j != s.positive_features.end(); j++) { + for (vector::const_iterator k = _feature2mef[*j].begin(); + k != _feature2mef[*j].end(); k++) { + update_folos_lazy(iter_sample, *k, _vl, sum_eta, last_updated); + } + } +#endif + + vector membp(_num_classes); + const int max_label = conditional_probability(s, membp); + + const double eta = + eta0 * pow(SGD_ALPHA, + (double)iter_sample / _vs.size()); // exponential decay + // const double eta = eta0 / (1.0 + (double)iter_sample / + // _vs.size()); + + // if (iter_sample % _vs.size() == 0) cerr << "eta = " << eta << + // endl; + u += eta * l1param; + + sum_eta.push_back(sum_eta.back() + eta * l1param); + + logl += log(membp[s.label]); + if (max_label == s.label) ncorrect++; + + // binary features + for (vector::const_iterator j = s.positive_features.begin(); + j != s.positive_features.end(); j++) { + for (vector::const_iterator k = _feature2mef[*j].begin(); + k != _feature2mef[*j].end(); k++) { + const double me = membp[_fb.Feature(*k).label()]; + const double ee = (_fb.Feature(*k).label() == s.label ? 1.0 : 0); + const double grad = (me - ee); + _vl[*k] -= eta * grad; +#ifdef SGD_CP + apply_l1_penalty(*k, u, _vl, q); +#endif + } + } + // real-valued features + for (vector >::const_iterator j = s.rvfeatures.begin(); + j != s.rvfeatures.end(); j++) { + for (vector::const_iterator k = _feature2mef[j->first].begin(); + k != _feature2mef[j->first].end(); k++) { + const double me = membp[_fb.Feature(*k).label()]; + const double ee = (_fb.Feature(*k).label() == s.label ? 1.0 : 0); + const double grad = (me - ee) * j->second; + _vl[*k] -= eta * grad; +#ifdef SGD_CP + apply_l1_penalty(*k, u, _vl, q); +#endif + } + } + +#ifdef FOLOS_NAIVE + for (size_t j = 0; j < d; j++) { + double& x = _vl[j]; + if (x > 0) + x = max(0.0, x - eta * l1param); + else + x = min(0.0, x + eta * l1param); + } +#endif + } + logl /= _vs.size(); +// fprintf(stderr, "%4d logl = %8.3f acc = %6.4f ", iter, logl, +// (double)ncorrect / ntotal); + +#ifdef FOLOS_LAZY + if (l1param > 0) { + for (size_t j = 0; j < d; j++) + update_folos_lazy(iter_sample, j, _vl, sum_eta, last_updated); + } +#endif + + double f = logl; + if (l1param > 0) { + const double l1 = + l1norm(_vl); // this is not accurate when lazy update is used + // cerr << "f0 = " << update_model_expectation() - l1param * l1 << " + // "; + f -= l1param * l1; + int nonzero = 0; + for (int j = 0; j < d; j++) + if (_vl[j] != 0) nonzero++; + // cerr << " f = " << f << " l1 = " << l1 << " nonzero_features = " + // << nonzero << endl; + } + // fprintf(stderr, "%4d obj = %7.3f acc = %6.4f", iter+1, f, + // (double)ncorrect/ntotal); + // fprintf(stderr, "%4d obj = %f", iter+1, f); + fprintf(stderr, "%3d obj(err) = %f (%6.4f)", iter + 1, f, + 1 - (double)ncorrect / ntotal); + + if (_nheldout > 0) { + double heldout_logl = heldout_likelihood(); + // fprintf(stderr, " heldout_logl = %f acc = %6.4f\n", + // heldout_logl, 1 - _heldout_error); + fprintf(stderr, " heldout_logl(err) = %f (%6.4f)", heldout_logl, + _heldout_error); + } + fprintf(stderr, "\n"); + } + + return 0; +} + +} // namespace maxent + /* * $Log: maxent.cpp,v $ * Revision 1.1.1.1 2007/05/15 08:30:35 kyoshida diff --git a/utils/maxent.h b/utils/maxent.h index b1efd88e..74d13a6f 100644 --- a/utils/maxent.h +++ b/utils/maxent.h @@ -5,21 +5,95 @@ #ifndef __MAXENT_H_ #define __MAXENT_H_ -#include -#include -#include -#include #include #include +#include +#include #include +#include +#include + #include -#include "mathvec.h" -#define USE_HASH_MAP // if you encounter errors with hash, try commenting out - // this line. (the program will be a bit slower, though) -#ifdef USE_HASH_MAP -#include -#endif +namespace maxent { +class Vec { + private: + std::vector _v; + + public: + Vec(const size_t n = 0, const double val = 0) { _v.resize(n, val); } + Vec(const std::vector& v) : _v(v) {} + const std::vector& STLVec() const { return _v; } + std::vector& STLVec() { return _v; } + size_t Size() const { return _v.size(); } + double& operator[](int i) { return _v[i]; } + const double& operator[](int i) const { return _v[i]; } + Vec& operator+=(const Vec& b) { + assert(b.Size() == _v.size()); + for (size_t i = 0; i < _v.size(); i++) { + _v[i] += b[i]; + } + return *this; + } + Vec& operator*=(const double c) { + for (size_t i = 0; i < _v.size(); i++) { + _v[i] *= c; + } + return *this; + } + void Project(const Vec& y) { + for (size_t i = 0; i < _v.size(); i++) { + // if (sign(_v[i]) != sign(y[i])) _v[i] = 0; + if (_v[i] * y[i] <= 0) _v[i] = 0; + } + } +}; + +inline double dot_product(const Vec& a, const Vec& b) { + double sum = 0; + for (size_t i = 0; i < a.Size(); i++) { + sum += a[i] * b[i]; + } + return sum; +} + +inline std::ostream& operator<<(std::ostream& s, const Vec& a) { + s << "("; + for (size_t i = 0; i < a.Size(); i++) { + if (i != 0) s << ", "; + s << a[i]; + } + s << ")"; + return s; +} + +inline const Vec operator+(const Vec& a, const Vec& b) { + Vec v(a.Size()); + assert(a.Size() == b.Size()); + for (size_t i = 0; i < a.Size(); i++) { + v[i] = a[i] + b[i]; + } + return v; +} + +inline const Vec operator-(const Vec& a, const Vec& b) { + Vec v(a.Size()); + assert(a.Size() == b.Size()); + for (size_t i = 0; i < a.Size(); i++) { + v[i] = a[i] - b[i]; + } + return v; +} + +inline const Vec operator*(const Vec& a, const double c) { + Vec v(a.Size()); + for (size_t i = 0; i < a.Size(); i++) { + v[i] = a[i] * c; + } + return v; +} + +inline const Vec operator*(const double c, const Vec& a) { return a * c; } // // data format for each sample for training/testing @@ -309,6 +383,7 @@ class ME_Model { static double FunctionGradientWrapper(const std::vector& x, std::vector& grad); }; +} // namespace maxent #endif diff --git a/utils/owlqn.cpp b/utils/owlqn.cpp deleted file mode 100644 index c3a0f0da..00000000 --- a/utils/owlqn.cpp +++ /dev/null @@ -1,127 +0,0 @@ -#include -#include -#include -#include -#include "mathvec.h" -#include "lbfgs.h" -#include "maxent.h" - -using namespace std; - -const static int M = LBFGS_M; -const static double LINE_SEARCH_ALPHA = 0.1; -const static double LINE_SEARCH_BETA = 0.5; - -// stopping criteria -int OWLQN_MAX_ITER = 300; -const static double MIN_GRAD_NORM = 0.0001; - -Vec approximate_Hg(const int iter, const Vec& grad, const Vec s[], - const Vec y[], const double z[]); - -inline int sign(double x) { - if (x > 0) return 1; - if (x < 0) return -1; - return 0; -}; - -static Vec pseudo_gradient(const Vec& x, const Vec& grad0, const double C) { - Vec grad = grad0; - for (size_t i = 0; i < x.Size(); i++) { - if (x[i] != 0) { - grad[i] += C * sign(x[i]); - continue; - } - const double gm = grad0[i] - C; - if (gm > 0) { - grad[i] = gm; - continue; - } - const double gp = grad0[i] + C; - if (gp < 0) { - grad[i] = gp; - continue; - } - grad[i] = 0; - } - - return grad; -} - -double ME_Model::regularized_func_grad(const double C, const Vec& x, - Vec& grad) { - double f = FunctionGradient(x.STLVec(), grad.STLVec()); - for (size_t i = 0; i < x.Size(); i++) { - f += C * fabs(x[i]); - } - - return f; -} - -double ME_Model::constrained_line_search(double C, const Vec& x0, - const Vec& grad0, const double f0, - const Vec& dx, Vec& x, Vec& grad1) { - // compute the orthant to explore - Vec orthant = x0; - for (size_t i = 0; i < orthant.Size(); i++) { - if (orthant[i] == 0) orthant[i] = -grad0[i]; - } - - double t = 1.0 / LINE_SEARCH_BETA; - - double f; - do { - t *= LINE_SEARCH_BETA; - x = x0 + t * dx; - x.Project(orthant); - // for (size_t i = 0; i < x.Size(); i++) { - // if (x0[i] != 0 && sign(x[i]) != sign(x0[i])) x[i] = 0; - // } - - f = regularized_func_grad(C, x, grad1); - // cout << "*"; - } while (f > f0 + LINE_SEARCH_ALPHA * dot_product(x - x0, grad0)); - - return f; -} - -vector ME_Model::perform_OWLQN(const vector& x0, - const double C) { - const size_t dim = x0.size(); - Vec x = x0; - - Vec grad(dim), dx(dim); - double f = regularized_func_grad(C, x, grad); - - Vec s[M], y[M]; - double z[M]; // rho - - for (int iter = 0; iter < OWLQN_MAX_ITER; iter++) { - Vec pg = pseudo_gradient(x, grad, C); - - fprintf(stderr, "%3d obj(err) = %f (%6.4f)", iter + 1, -f, _train_error); - if (_nheldout > 0) { - const double heldout_logl = heldout_likelihood(); - fprintf(stderr, " heldout_logl(err) = %f (%6.4f)", heldout_logl, - _heldout_error); - } - fprintf(stderr, "\n"); - - if (sqrt(dot_product(pg, pg)) < MIN_GRAD_NORM) break; - - dx = -1 * approximate_Hg(iter, pg, s, y, z); - if (dot_product(dx, pg) >= 0) dx.Project(-1 * pg); - - Vec x1(dim), grad1(dim); - f = constrained_line_search(C, x, pg, f, dx, x1, grad1); - - s[iter % M] = x1 - x; - y[iter % M] = grad1 - grad; - z[iter % M] = 1.0 / dot_product(y[iter % M], s[iter % M]); - - x = x1; - grad = grad1; - } - - return x.STLVec(); -} diff --git a/utils/sgd.cpp b/utils/sgd.cpp deleted file mode 100644 index 8613edca..00000000 --- a/utils/sgd.cpp +++ /dev/null @@ -1,193 +0,0 @@ -#include "maxent.h" -#include -#include - -using namespace std; - -// const double SGD_ETA0 = 1; -// const double SGD_ITER = 30; -// const double SGD_ALPHA = 0.85; - -//#define FOLOS_NAIVE -//#define FOLOS_LAZY -#define SGD_CP - -inline void apply_l1_penalty(const int i, const double u, vector& _vl, - vector& q) { - double& w = _vl[i]; - const double z = w; - double& qi = q[i]; - if (w > 0) { - w = max(0.0, w - (u + qi)); - } else if (w < 0) { - w = min(0.0, w + (u - qi)); - } - qi += w - z; -} - -static double l1norm(const vector& v) { - double sum = 0; - for (size_t i = 0; i < v.size(); i++) sum += abs(v[i]); - return sum; -} - -inline void update_folos_lazy(const int iter_sample, const int k, - vector& _vl, - const vector& sum_eta, - vector& last_updated) { - const double penalty = sum_eta[iter_sample] - sum_eta[last_updated[k]]; - double& x = _vl[k]; - if (x > 0) - x = max(0.0, x - penalty); - else - x = min(0.0, x + penalty); - last_updated[k] = iter_sample; -} - -int ME_Model::perform_SGD() { - if (_l2reg > 0) { - cerr << "error: L2 regularization is currently not supported in SGD mode." - << endl; - exit(1); - } - - cerr << "performing SGD" << endl; - - const double l1param = _l1reg; - - const int d = _fb.Size(); - - vector ri(_vs.size()); - for (size_t i = 0; i < ri.size(); i++) ri[i] = i; - - vector grad(d); - int iter_sample = 0; - const double eta0 = SGD_ETA0; - - // cerr << "l1param = " << l1param << endl; - cerr << "eta0 = " << eta0 << " alpha = " << SGD_ALPHA << endl; - - double u = 0; - vector q(d, 0); - vector last_updated(d, 0); - vector sum_eta; - sum_eta.push_back(0); - - for (int iter = 0; iter < SGD_ITER; iter++) { - - random_shuffle(ri.begin(), ri.end()); - - double logl = 0; - int ncorrect = 0, ntotal = 0; - for (size_t i = 0; i < _vs.size(); i++, ntotal++, iter_sample++) { - const Sample& s = _vs[ri[i]]; - -#ifdef FOLOS_LAZY - for (vector::const_iterator j = s.positive_features.begin(); - j != s.positive_features.end(); j++) { - for (vector::const_iterator k = _feature2mef[*j].begin(); - k != _feature2mef[*j].end(); k++) { - update_folos_lazy(iter_sample, *k, _vl, sum_eta, last_updated); - } - } -#endif - - vector membp(_num_classes); - const int max_label = conditional_probability(s, membp); - - const double eta = - eta0 * pow(SGD_ALPHA, - (double)iter_sample / _vs.size()); // exponential decay - // const double eta = eta0 / (1.0 + (double)iter_sample / - // _vs.size()); - - // if (iter_sample % _vs.size() == 0) cerr << "eta = " << eta << - // endl; - u += eta * l1param; - - sum_eta.push_back(sum_eta.back() + eta * l1param); - - logl += log(membp[s.label]); - if (max_label == s.label) ncorrect++; - - // binary features - for (vector::const_iterator j = s.positive_features.begin(); - j != s.positive_features.end(); j++) { - for (vector::const_iterator k = _feature2mef[*j].begin(); - k != _feature2mef[*j].end(); k++) { - const double me = membp[_fb.Feature(*k).label()]; - const double ee = (_fb.Feature(*k).label() == s.label ? 1.0 : 0); - const double grad = (me - ee); - _vl[*k] -= eta * grad; -#ifdef SGD_CP - apply_l1_penalty(*k, u, _vl, q); -#endif - } - } - // real-valued features - for (vector >::const_iterator j = s.rvfeatures.begin(); - j != s.rvfeatures.end(); j++) { - for (vector::const_iterator k = _feature2mef[j->first].begin(); - k != _feature2mef[j->first].end(); k++) { - const double me = membp[_fb.Feature(*k).label()]; - const double ee = (_fb.Feature(*k).label() == s.label ? 1.0 : 0); - const double grad = (me - ee) * j->second; - _vl[*k] -= eta * grad; -#ifdef SGD_CP - apply_l1_penalty(*k, u, _vl, q); -#endif - } - } - -#ifdef FOLOS_NAIVE - for (size_t j = 0; j < d; j++) { - double& x = _vl[j]; - if (x > 0) - x = max(0.0, x - eta * l1param); - else - x = min(0.0, x + eta * l1param); - } -#endif - } - logl /= _vs.size(); -// fprintf(stderr, "%4d logl = %8.3f acc = %6.4f ", iter, logl, -// (double)ncorrect / ntotal); - -#ifdef FOLOS_LAZY - if (l1param > 0) { - for (size_t j = 0; j < d; j++) - update_folos_lazy(iter_sample, j, _vl, sum_eta, last_updated); - } -#endif - - double f = logl; - if (l1param > 0) { - const double l1 = - l1norm(_vl); // this is not accurate when lazy update is used - // cerr << "f0 = " << update_model_expectation() - l1param * l1 << " - // "; - f -= l1param * l1; - int nonzero = 0; - for (int j = 0; j < d; j++) - if (_vl[j] != 0) nonzero++; - // cerr << " f = " << f << " l1 = " << l1 << " nonzero_features = " - // << nonzero << endl; - } - // fprintf(stderr, "%4d obj = %7.3f acc = %6.4f", iter+1, f, - // (double)ncorrect/ntotal); - // fprintf(stderr, "%4d obj = %f", iter+1, f); - fprintf(stderr, "%3d obj(err) = %f (%6.4f)", iter + 1, f, - 1 - (double)ncorrect / ntotal); - - if (_nheldout > 0) { - double heldout_logl = heldout_likelihood(); - // fprintf(stderr, " heldout_logl = %f acc = %6.4f\n", - // heldout_logl, 1 - _heldout_error); - fprintf(stderr, " heldout_logl(err) = %f (%6.4f)", heldout_logl, - _heldout_error); - } - fprintf(stderr, "\n"); - } - - return 0; -} -- cgit v1.2.3