Merge pull request #11 from bab2min/develop

Develop

README.kr.rst

@@ -28,7 +28,7 @@ tomotopy 란?
 
 더 자세한 정보는 https://bab2min.github.io/tomotopy/index.kr.html 에서 확인하시길 바랍니다.
 
-tomotopy의 가장 최신버전은 0.3.0 입니다.
+tomotopy의 가장 최신버전은 0.3.1 입니다.
 
 시작하기
 ---------------
@@ -195,6 +195,11 @@ tomotopy의 Python3 예제 코드는 https://github.com/bab2min/tomotopy/blob/ma
 
 역사
 -------
+* 0.3.1 (2019-11-05)
+    * `min_cf` 혹은 `rm_top`가 설정되었을 때 `get_topic_dist()`의 반환값이 부정확한 문제를 수정하였습니다.
+    * `tomotopy.MGLDAModel` 모델의 문헌의 `get_topic_dist()`가 지역 토픽에 대한 분포도 함께 반환하도록 수정하였습니다..
+    * `tw=ONE`일때의 학습 속도가 개선되었습니다.
+
 * 0.3.0 (2019-10-06)
     * `tomotopy.LLDAModel` 토픽 모델이 새로 추가되었습니다.
     * `HDPModel`을 학습할 때 프로그램이 종료되는 문제를 해결했습니다.

README.rst

@@ -29,7 +29,7 @@ The current version of `tomoto` supports several major topic models including
 
 Please visit https://bab2min.github.io/tomotopy to see more information.
 
-The most recent version of tomotopy is 0.3.0.
+The most recent version of tomotopy is 0.3.1.
 
 Getting Started
 ---------------
@@ -200,6 +200,11 @@ meaning you can use it for any reasonable purpose and remain in complete ownersh
 
 History
 -------
+* 0.3.1 (2019-11-05)
+    * An issue where `get_topic_dist()` returns incorrect value when `min_cf` or `rm_top` is set was fixed.
+    * The return value of `get_topic_dist()` of `tomotopy.MGLDAModel` document was fixed to include local topics.
+    * The estimation speed with `tw=ONE` was improved.
+
 * 0.3.0 (2019-10-06)
     * A new model, `tomotopy.LLDAModel` was added into the package.
     * A crashing issue of `HDPModel` was fixed.

setup.py

@@ -24,8 +24,8 @@
     largs += ['-stdlib=libc++']
     arch_levels = {'':'-march=native'}
 elif 'manylinux' in os.environ.get('AUDITWHEEL_PLAT', ''):
-    cargs = ['-std=c++0x', '-O3', '-fpermissive']
-    arch_levels = {'':'', 'sse2':'-msse2', 'avx':'-mavx'}
+    cargs = ['-std=c++0x', '-O3', '-fpermissive', '-g0']
+    arch_levels = {'':'', 'sse2':'-msse2', 'avx':'-mavx', 'avx2':'-mavx2'}
 else:
     cargs = ['-std=c++0x', '-O3', '-fpermissive']
     arch_levels = {'':'-march=native'}
@@ -47,7 +47,7 @@
 setup(
     name='tomotopy',
 
-    version='0.3.0',
+    version='0.3.1',
 
     description='Tomoto, The Topic Modeling Tool for Python',
     long_description=long_description,

src/TopicModel/CT.h

@@ -3,13 +3,13 @@
 
 namespace tomoto
 {
-    template<TermWeight _TW, bool _Shared = false>
-	struct DocumentCTM : public DocumentLDA<_TW, _Shared>
+    template<TermWeight _TW, size_t _Flags = 0>
+	struct DocumentCTM : public DocumentLDA<_TW, _Flags>
 	{
-		using DocumentLDA<_TW, _Shared>::DocumentLDA;
+		using DocumentLDA<_TW, _Flags>::DocumentLDA;
 		Eigen::Matrix<FLOAT, -1, -1> beta; // Dim: (K, betaSample)
 		Eigen::Matrix<FLOAT, -1, 1> smBeta; // Dim: K
-		DEFINE_SERIALIZER_AFTER_BASE2(DocumentLDA<_TW, _Shared>, smBeta);
+		DEFINE_SERIALIZER_AFTER_BASE2(DocumentLDA<_TW, _Flags>, smBeta);
 	};
 
 	class ICTModel : public ILDAModel
@@ -18,7 +18,7 @@ namespace tomoto
 		using DefaultDocType = DocumentCTM<TermWeight::one>;
 		static ICTModel* create(TermWeight _weight, size_t _K = 1,
 			FLOAT smoothingAlpha = 0.1,  FLOAT _eta = 0.01,
-			const RANDGEN& _rg = RANDGEN{ std::random_device{}() });
+			const RandGen& _rg = RandGen{ std::random_device{}() });
 
 		virtual void setNumBetaSample(size_t numSample) = 0;
 		virtual size_t getNumBetaSample() const = 0;

src/TopicModel/CTModel.cpp

@@ -6,7 +6,7 @@ namespace tomoto
 	template class CTModel<TermWeight::idf>;
 	template class CTModel<TermWeight::pmi>;
 
-	ICTModel* ICTModel::create(TermWeight _weight, size_t _K, FLOAT smoothingAlpha, FLOAT _eta, const RANDGEN& _rg)
+	ICTModel* ICTModel::create(TermWeight _weight, size_t _K, FLOAT smoothingAlpha, FLOAT _eta, const RandGen& _rg)
 	{
 		SWITCH_TW(_weight, CTModel, _K, smoothingAlpha, _eta, _rg);
 	}

src/TopicModel/CTModel.hpp

@@ -16,19 +16,19 @@ namespace tomoto
 	{
 	};
 
-	template<TermWeight _TW, bool _Shared = false,
+	template<TermWeight _TW, size_t _Flags = 0,
 		typename _Interface = ICTModel,
 		typename _Derived = void,
 		typename _DocType = DocumentCTM<_TW>,
 		typename _ModelState = ModelStateCTM<_TW>>
-		class CTModel : public LDAModel<_TW, _Shared, _Interface,
-		typename std::conditional<std::is_same<_Derived, void>::value, CTModel<_TW, _Shared>, _Derived>::type,
+	class CTModel : public LDAModel<_TW, _Flags, _Interface,
+		typename std::conditional<std::is_same<_Derived, void>::value, CTModel<_TW, _Flags>, _Derived>::type,
 		_DocType, _ModelState>
 	{
 		static constexpr const char* TMID = "CTM";
 	protected:
 		using DerivedClass = typename std::conditional<std::is_same<_Derived, void>::value, CTModel<_TW>, _Derived>::type;
-		using BaseClass = LDAModel<_TW, _Shared, _Interface, DerivedClass, _DocType, _ModelState>;
+		using BaseClass = LDAModel<_TW, _Flags, _Interface, DerivedClass, _DocType, _ModelState>;
 		friend BaseClass;
 		friend typename BaseClass::BaseClass;
 		using WeightType = typename BaseClass::WeightType;
@@ -50,7 +50,7 @@ namespace tomoto
 			return &zLikelihood[0];
 		}
 
-		void updateBeta(_DocType& doc, RANDGEN& rg) const
+		void updateBeta(_DocType& doc, RandGen& rg) const
 		{
 			Eigen::Matrix<FLOAT, -1, 1> pbeta, lowerBound, upperBound;
 			constexpr FLOAT epsilon = 1e-8;
@@ -65,7 +65,7 @@ namespace tomoto
 				for (size_t k = 0; k < this->K; ++k)
 				{
 					FLOAT N_k = doc.numByTopic[k] + this->alpha;
-					FLOAT N_nk = doc.template getSumWordWeight<_TW>() + this->alpha * (this->K + 1) - N_k;
+					FLOAT N_nk = doc.getSumWordWeight() + this->alpha * (this->K + 1) - N_k;
 					FLOAT u1 = std::generate_canonical<FLOAT, 32>(rg), u2 = std::generate_canonical<FLOAT, 32>(rg);
 					FLOAT max_uk = epsilon + pow(u1, (FLOAT)1 / N_k)  * (pbeta[k] - epsilon);
 					FLOAT min_unk = (1 - pow(u2, (FLOAT)1 / N_nk))
@@ -104,7 +104,7 @@ namespace tomoto
 			doc.smBeta /= doc.smBeta.array().sum();
 		}
 
-		void sampleDocument(_DocType& doc, size_t docId, _ModelState& ld, RANDGEN& rgs, size_t iterationCnt) const
+		void sampleDocument(_DocType& doc, size_t docId, _ModelState& ld, RandGen& rgs, size_t iterationCnt) const
 		{
 			BaseClass::sampleDocument(doc, docId, ld, rgs, iterationCnt);
 			if (iterationCnt >= this->burnIn && this->optimInterval && (iterationCnt + 1) % this->optimInterval == 0)
@@ -113,7 +113,7 @@ namespace tomoto
 			}
 		}
 
-		int restoreFromTrainingError(const exception::TrainingError& e, ThreadPool& pool, _ModelState* localData, RANDGEN* rgs)
+		int restoreFromTrainingError(const exception::TrainingError& e, ThreadPool& pool, _ModelState* localData, RandGen* rgs)
 		{
 			std::cerr << "Failed to sample! Reset prior and retry!" << std::endl;
 			const size_t chStride = std::min(pool.getNumWorkers() * 8, this->docs.size());
@@ -134,7 +134,7 @@ namespace tomoto
 			return 0;
 		}
 
-		void optimizeParameters(ThreadPool& pool, _ModelState* localData, RANDGEN* rgs)
+		void optimizeParameters(ThreadPool& pool, _ModelState* localData, RandGen* rgs)
 		{
 			std::vector<std::future<void>> res;
 			topicPrior = math::MultiNormalDistribution<FLOAT>::estimate([this](size_t i)
@@ -164,7 +164,7 @@ namespace tomoto
 				}
 				pbeta.array() -= last;
 				ll += topicPrior.getLL(pbeta.head(this->K));
-				ll += math::lgammaT(doc.template getSumWordWeight<_TW>() + alpha * K + 1);
+				ll += math::lgammaT(doc.getSumWordWeight() + alpha * K + 1);
 			}
 			return ll;
 		}
@@ -197,7 +197,7 @@ namespace tomoto
 		DEFINE_SERIALIZER_AFTER_BASE(BaseClass, numBetaSample, numTMNSample, topicPrior);
 
 	public:
-		CTModel(size_t _K = 1, FLOAT smoothingAlpha = 0.1, FLOAT _eta = 0.01, const RANDGEN& _rg = RANDGEN{ std::random_device{}() })
+		CTModel(size_t _K = 1, FLOAT smoothingAlpha = 0.1, FLOAT _eta = 0.01, const RandGen& _rg = RandGen{ std::random_device{}() })
 			: BaseClass(_K, smoothingAlpha, _eta, _rg)
 		{
 			this->optimInterval = 2;
@@ -206,11 +206,8 @@ namespace tomoto
 		std::vector<FLOAT> getTopicsByDoc(const _DocType& doc) const
 		{
 			std::vector<FLOAT> ret(this->K);
-			FLOAT sum = doc.template getSumWordWeight<_TW>();
-			for (size_t k = 0; k < this->K; ++k)
-			{
-				ret[k] = doc.numByTopic[k] / sum;
-			}
+			Eigen::Map<Eigen::Matrix<FLOAT, -1, 1>>{ret.data(), this->K}.array() =
+				doc.numByTopic.array().template cast<FLOAT>() / doc.getSumWordWeight();
 			return ret;
 		}
 

src/TopicModel/DMR.h

@@ -3,13 +3,13 @@
 
 namespace tomoto
 {
-	template<TermWeight _TW, bool _Shared = false>
-	struct DocumentDMR : public DocumentLDA<_TW, _Shared>
+	template<TermWeight _TW, size_t _Flags = 0>
+	struct DocumentDMR : public DocumentLDA<_TW, _Flags>
 	{
-		using DocumentLDA<_TW, _Shared>::DocumentLDA;
+		using DocumentLDA<_TW, _Flags>::DocumentLDA;
 		size_t metadata = 0;
 
-		DEFINE_SERIALIZER_AFTER_BASE2(DocumentLDA<_TW, _Shared>, metadata);
+		DEFINE_SERIALIZER_AFTER_BASE2(DocumentLDA<_TW, _Flags>, metadata);
 	};
 
     class IDMRModel : public ILDAModel
@@ -18,7 +18,7 @@ namespace tomoto
 		using DefaultDocType = DocumentDMR<TermWeight::one>;
 		static IDMRModel* create(TermWeight _weight, size_t _K = 1,
 			FLOAT defaultAlpha = 1.0, FLOAT _sigma = 1.0, FLOAT _eta = 0.01, FLOAT _alphaEps = 1e-10,
-			const RANDGEN& _rg = RANDGEN{ std::random_device{}() });
+			const RandGen& _rg = RandGen{ std::random_device{}() });
 
 		virtual size_t addDoc(const std::vector<std::string>& words, const std::vector<std::string>& metadata) = 0;
 		virtual std::unique_ptr<DocumentBase> makeDoc(const std::vector<std::string>& words, const std::vector<std::string>& metadata) const = 0;

src/TopicModel/DMRModel.cpp

@@ -6,7 +6,7 @@ namespace tomoto
 	template class DMRModel<TermWeight::idf>;
 	template class DMRModel<TermWeight::pmi>;
 
-	IDMRModel* IDMRModel::create(TermWeight _weight, size_t _K, FLOAT _defaultAlpha, FLOAT _sigma, FLOAT _eta, FLOAT _alphaEps, const RANDGEN& _rg)
+	IDMRModel* IDMRModel::create(TermWeight _weight, size_t _K, FLOAT _defaultAlpha, FLOAT _sigma, FLOAT _eta, FLOAT _alphaEps, const RandGen& _rg)
 	{
 		SWITCH_TW(_weight, DMRModel, _K, _defaultAlpha, _sigma, _eta, _alphaEps, _rg);
 	}

src/TopicModel/DMRModel.hpp

@@ -16,19 +16,19 @@ namespace tomoto
 		Eigen::Matrix<FLOAT, -1, 1> tmpK;
 	};
 
-	template<TermWeight _TW, bool _Shared = false,
+	template<TermWeight _TW, size_t _Flags = 0,
 		typename _Interface = IDMRModel,
 		typename _Derived = void,
 		typename _DocType = DocumentDMR<_TW>,
 		typename _ModelState = ModelStateDMR<_TW>>
-	class DMRModel : public LDAModel<_TW, _Shared, _Interface,
-		typename std::conditional<std::is_same<_Derived, void>::value, DMRModel<_TW, _Shared>, _Derived>::type,
+	class DMRModel : public LDAModel<_TW, _Flags, _Interface,
+		typename std::conditional<std::is_same<_Derived, void>::value, DMRModel<_TW, _Flags>, _Derived>::type,
 		_DocType, _ModelState>
 	{
 		static constexpr const char* TMID = "DMR";
 	protected:
 		using DerivedClass = typename std::conditional<std::is_same<_Derived, void>::value, DMRModel<_TW>, _Derived>::type;
-		using BaseClass = LDAModel<_TW, _Shared, _Interface, DerivedClass, _DocType, _ModelState>;
+		using BaseClass = LDAModel<_TW, _Flags, _Interface, DerivedClass, _DocType, _ModelState>;
 		friend BaseClass;
 		friend typename BaseClass::BaseClass;
 		using WeightType = typename BaseClass::WeightType;
@@ -83,8 +83,8 @@ namespace tomoto
 						}
 						//val[K * F] = -(lgammaApprox(alphaDoc.array()) - lgammaApprox(doc.numByTopic.array().cast<FLOAT>() + alphaDoc.array())).sum();
 						//tmpK = -(digammaApprox(alphaDoc.array()) - digammaApprox(doc.numByTopic.array().cast<FLOAT>() + alphaDoc.array()));
-						val[K * F] += math::lgammaT(alphaSum) - math::lgammaT(doc.template getSumWordWeight<_TW>() + alphaSum);
-						FLOAT t = math::digammaT(alphaSum) - math::digammaT(doc.template getSumWordWeight<_TW>() + alphaSum);
+						val[K * F] += math::lgammaT(alphaSum) - math::lgammaT(doc.getSumWordWeight() + alphaSum);
+						FLOAT t = math::digammaT(alphaSum) - math::digammaT(doc.getSumWordWeight() + alphaSum);
 						if (!std::isfinite(alphaSum) && alphaSum > 0)
 						{
 							val[K * F] = -INFINITY;
@@ -116,7 +116,7 @@ namespace tomoto
 			}
 		}
 
-		void optimizeParameters(ThreadPool& pool, _ModelState* localData, RANDGEN* rgs)
+		void optimizeParameters(ThreadPool& pool, _ModelState* localData, RandGen* rgs)
 		{
 			Eigen::Matrix<FLOAT, -1, -1> bLambda;
 			FLOAT fx = 0, bestFx = INFINITY;
@@ -137,14 +137,21 @@ namespace tomoto
 			}
 			if (!std::isfinite(bestFx))
 			{
-				std::cout << "optimizing parameters has been failed!" << std::endl;
-				throw std::runtime_error{ "optimizing parameters has been failed!" };
+				throw exception::TrainingError{ "optimizing parameters has been failed!" };
 			}
 			lambda = bLambda;
 			//std::cerr << fx << std::endl;
 			expLambda = lambda.array().exp() + alphaEps;
 		}
 
+		int restoreFromTrainingError(const exception::TrainingError& e, ThreadPool& pool, _ModelState* localData, RandGen* rgs)
+		{
+			std::cerr << "Failed to optimize! Reset prior and retry!" << std::endl;
+			lambda.setZero();
+			expLambda = lambda.array().exp() + alphaEps;
+			return 0;
+		}
+
 		FLOAT* getZLikelihoods(_ModelState& ld, const _DocType& doc, size_t docId, size_t vid) const
 		{
 			const size_t V = this->realV;
@@ -173,7 +180,7 @@ namespace tomoto
 				ll += math::lgammaT(doc.numByTopic[k] + alphaDoc[k]);
 				ll -= math::lgammaT(alphaDoc[k]);
 			}
-			ll -= math::lgammaT(doc.template getSumWordWeight<_TW>() + alphaSum);
+			ll -= math::lgammaT(doc.getSumWordWeight() + alphaSum);
 			ll += math::lgammaT(alphaSum);
 			return ll;
 		}
@@ -194,7 +201,7 @@ namespace tomoto
 				{
 					ll += math::lgammaT(doc.numByTopic[k] + alphaDoc[k]) - math::lgammaT(alphaDoc[k]);
 				}
-				ll -= math::lgammaT(doc.template getSumWordWeight<_TW>() + alphaSum) - math::lgammaT(alphaSum);
+				ll -= math::lgammaT(doc.getSumWordWeight() + alphaSum) - math::lgammaT(alphaSum);
 			}
 			return ll;
 		}
@@ -231,7 +238,7 @@ namespace tomoto
 			{
 				lambda = Eigen::Matrix<FLOAT, -1, -1>::Constant(this->K, F, log(this->alpha));
 			}
-			if (_Shared) this->numByTopicDoc = Eigen::Matrix<WeightType, -1, -1>::Zero(this->K, this->docs.size());
+			if (_Flags & flags::continuous_doc_data) this->numByTopicDoc = Eigen::Matrix<WeightType, -1, -1>::Zero(this->K, this->docs.size());
 			expLambda = lambda.array().exp();
 			LBFGSpp::LBFGSParam<FLOAT> param;
 			param.max_iterations = maxBFGSIteration;
@@ -242,7 +249,7 @@ namespace tomoto
 
 	public:
 		DMRModel(size_t _K = 1, FLOAT defaultAlpha = 1.0, FLOAT _sigma = 1.0, FLOAT _eta = 0.01, 
-			FLOAT _alphaEps = 0, const RANDGEN& _rg = RANDGEN{ std::random_device{}() })
+			FLOAT _alphaEps = 0, const RandGen& _rg = RandGen{ std::random_device{}() })
 			: BaseClass(_K, defaultAlpha, _eta, _rg), sigma(_sigma), alphaEps(_alphaEps)
 		{
 		}
@@ -285,11 +292,8 @@ namespace tomoto
 		{
 			std::vector<FLOAT> ret(this->K);
 			auto alphaDoc = expLambda.col(doc.metadata);
-			FLOAT sum = doc.template getSumWordWeight<_TW>() + alphaDoc.sum();
-			for (size_t k = 0; k < this->K; ++k)
-			{
-				ret[k] = (doc.numByTopic[k] + alphaDoc[k]) / sum;
-			}
+			Eigen::Map<Eigen::Matrix<FLOAT, -1, 1>>{ret.data(), this->K}.array() =
+				(doc.numByTopic.array().template cast<FLOAT>() + alphaDoc.array()) / (doc.getSumWordWeight() + alphaDoc.sum());
 			return ret;
 		}
 
@@ -311,11 +315,11 @@ namespace tomoto
 	};
 
 	/* This is for preventing 'undefined symbol' problem in compiling by clang. */
-	template<TermWeight _TW, bool _Shared,
+	template<TermWeight _TW, size_t _Flags,
 		typename _Interface, typename _Derived, typename _DocType, typename _ModelState>
-		constexpr FLOAT DMRModel<_TW, _Shared, _Interface, _Derived, _DocType, _ModelState>::maxLambda;
+		constexpr FLOAT DMRModel<_TW, _Flags, _Interface, _Derived, _DocType, _ModelState>::maxLambda;
 
-	template<TermWeight _TW, bool _Shared,
+	template<TermWeight _TW, size_t _Flags,
 		typename _Interface, typename _Derived, typename _DocType, typename _ModelState>
-		constexpr size_t DMRModel<_TW, _Shared, _Interface, _Derived, _DocType, _ModelState>::maxBFGSIteration;
+		constexpr size_t DMRModel<_TW, _Flags, _Interface, _Derived, _DocType, _ModelState>::maxBFGSIteration;
 }

src/TopicModel/GDMR.h

@@ -3,13 +3,13 @@
 
 namespace tomoto
 {
-    template<TermWeight _TW, bool _Shared = false>
-	struct DocumentGDMR : public DocumentDMR<_TW, _Shared>
+    template<TermWeight _TW, size_t _Flags = 0>
+	struct DocumentGDMR : public DocumentDMR<_TW, _Flags>
 	{
-		using DocumentDMR<_TW, _Shared>::DocumentDMR;
+		using DocumentDMR<_TW, _Flags>::DocumentDMR;
 		std::vector<FLOAT> metadataC;
 
-		DEFINE_SERIALIZER_AFTER_BASE2(DocumentDMR<_TW, _Shared>, metadataC);
+		DEFINE_SERIALIZER_AFTER_BASE2(DocumentDMR<_TW, _Flags>, metadataC);
 	};
 
     class IGDMRModel : public IDMRModel
@@ -18,7 +18,7 @@ namespace tomoto
 		using DefaultDocType = DocumentDMR<TermWeight::one>;
 		static IGDMRModel* create(TermWeight _weight, size_t _K = 1, const std::vector<size_t>& _degreeByF = {},
 			FLOAT defaultAlpha = 1.0, FLOAT _sigma = 1.0, FLOAT _eta = 0.01, FLOAT _alphaEps = 1e-10,
-			const RANDGEN& _rg = RANDGEN{ std::random_device{}() });
+			const RandGen& _rg = RandGen{ std::random_device{}() });
 
 		virtual FLOAT getSigma0() const = 0;
 		virtual void setSigma0(FLOAT) = 0;