Added Multitask L1/L2 logistic regression

shogun-toolbox · Jul 15, 2012 · 629358d · 629358d
1 parent fc2e96c
commit 629358d
Show file tree

Hide file tree

Showing 6 changed files with 397 additions and 0 deletions.
diff --git a/src/interfaces/modular/Transfer.i b/src/interfaces/modular/Transfer.i
@@ -20,6 +20,7 @@
 %rename(TaskTree) CTaskTree;
 %rename(MultitaskLSRegression) CMultitaskLSRegression;
 %rename(MultitaskLogisticRegression) CMultitaskLogisticRegression;
+%rename(MultitaskL1L2LogisticRegression) CMultitaskL1L2LogisticRegression;
 
 %rename(LibLinearMTL) CLibLinearMTL;
 
@@ -42,6 +43,7 @@
 %include <shogun/transfer/multitask/TaskTree.h>
 %include <shogun/transfer/multitask/MultitaskLSRegression.h>
 %include <shogun/transfer/multitask/MultitaskLogisticRegression.h>
+%include <shogun/transfer/multitask/MultitaskL1L2LogisticRegression.h>
 
 %include <shogun/transfer/multitask/LibLinearMTL.h>
 

diff --git a/src/interfaces/modular/Transfer_includes.i b/src/interfaces/modular/Transfer_includes.i
@@ -11,6 +11,7 @@
  #include <shogun/transfer/multitask/TaskTree.h>
  #include <shogun/transfer/multitask/MultitaskLSRegression.h>
  #include <shogun/transfer/multitask/MultitaskLogisticRegression.h>
+ #include <shogun/transfer/multitask/MultitaskL1L2LogisticRegression.h>
 
 #ifdef USE_SVMLIGHT
  #include <shogun/transfer/domain_adaptation/DomainAdaptationSVM.h>

diff --git a/src/shogun/lib/slep/malsar_joint_feature_learning.cpp b/src/shogun/lib/slep/malsar_joint_feature_learning.cpp
@@ -0,0 +1,215 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * Written (W) 2012 Sergey Lisitsyn
+ * Copyright (C) 2012 Jiayu Zhou and Jieping Ye
+ */
+
+#include <shogun/lib/slep/malsar_joint_feature_learning.h>
+#include <shogun/mathematics/Math.h>
+#include <iostream>
+
+#define EIGEN_RUNTIME_NO_MALLOC
+#define NDEBUG
+#include <eigen3/Eigen/Dense>
+
+using namespace Eigen;
+
+namespace shogun
+{
+
+slep_result_t malsar_joint_feature_learning(
+		CDotFeatures* features,
+		double* y,
+		double rho1,
+		double rho2,
+		const slep_options& options)
+{
+	int task;
+	int n_feats = features->get_dim_feature_space();
+	SG_SPRINT("n feats = %d\n", n_feats);
+	int n_vecs = features->get_num_vectors();
+	SG_SPRINT("n vecs = %d\n", n_vecs);
+	int n_tasks = options.n_tasks;
+	SG_SPRINT("n tasks = %d\n", n_tasks);
+
+	int iter = 0;
+
+	// initialize weight vector and bias for each task
+	MatrixXd Ws = MatrixXd::Zero(n_feats, n_tasks);
+	VectorXd Cs = VectorXd::Zero(n_tasks);
+	for (task=0; task<n_tasks; task++)
+	{
+		int n_pos = 0;
+		int n_neg = 0;
+		for (int i=options.ind[task]; i<options.ind[task+1]; i++)
+		{
+			if (y[i] > 0)
+				n_pos++;
+			else
+				n_neg++;
+		}
+		Cs[task] = CMath::log(double(n_pos)/n_neg);
+	}
+
+	MatrixXd Wz=Ws, Wzp=Ws, Wz_old=Ws, delta_Wzp=Ws, gWs=Ws;
+	VectorXd Cz=Cs, Czp=Cs, Cz_old=Cs, delta_Czp=Cs, gCs=Cs;
+
+	double t=1, t_old=0;
+	double gamma=1, gamma_inc=2;
+	double obj=0.0, obj_old=0.0;
+
+	internal::set_is_malloc_allowed(false);
+	while (iter < options.max_iter)
+	{
+		double alpha = double(t_old - 1)/t;
+
+		// compute search point
+		Ws = (1+alpha)*Wz - alpha*Wz_old;
+		Cs = (1+alpha)*Cz - alpha*Cz_old;
+
+		// zero gradient
+		gWs.setZero();
+		gCs.setZero();
+
+		// compute gradient and objective at search point
+		double Fs = 0;
+		for (task=0; task<n_tasks; task++)
+		{
+			for (int i=options.ind[task]; i<options.ind[task+1]; i++)
+			{
+				double aa = -y[i]*(features->dense_dot(i, Ws.col(task).data(), n_feats)+Cs[task]);
+				double bb = CMath::max(aa,0.0);
+
+				// avoid underflow when computing exponential loss
+				Fs += (CMath::log(CMath::exp(-bb) + CMath::exp(aa-bb)) + bb)/n_vecs;
+				double b = -y[i]*(1 - 1/(1+CMath::exp(aa)))/n_vecs;
+
+				gCs[task] += b;
+				features->add_to_dense_vec(b, i, gWs.col(task).data(), n_feats);
+			}
+		}
+		gWs.noalias() += 2*rho2*Ws;
+
+		// add regularizer
+		Fs += Ws.squaredNorm();
+
+		double Fzp = 0.0;
+		double gradient_break = false;
+
+		// line search, Armijo-Goldstein scheme
+		while (true)
+		{
+			// compute lasso projection of Ws - gWs/gamma
+			for (task=0; task<n_tasks; task++)
+			{
+				Wzp.col(task) = Ws.col(task) - gWs.col(task)/gamma;
+				double norm = Wzp.col(task).lpNorm<2>();
+				Wzp.col(task) *= CMath::max(0.0,norm-rho1/gamma)/norm;
+			}
+			// walk in direction of antigradient 
+			Czp = Cs - gCs/gamma;
+
+			// compute objective at line search point
+			Fzp = 0.0;
+			for (task=0; task<n_tasks; task++)
+			{
+				for (int i=options.ind[task]; i<options.ind[task+1]; i++)
+				{
+					double aa = -y[i]*(features->dense_dot(i, Wzp.col(task).data(), n_feats)+Cs[task]);
+					double bb = CMath::max(aa,0.0);
+
+					Fzp += (CMath::log(CMath::exp(-bb) + CMath::exp(aa-bb)) + bb)/n_vecs;
+				}
+			}
+			Fzp += Wzp.squaredNorm();
+
+			// compute delta between line search point and search point
+			delta_Wzp = Wzp - Ws;
+			delta_Czp = Czp - Cs;
+
+			// norms of delta
+			double nrm_delta_Wzp = delta_Wzp.squaredNorm();
+			double nrm_delta_Czp = delta_Czp.squaredNorm();
+
+			double r_sum = (nrm_delta_Wzp + nrm_delta_Czp)/2;
+
+			double Fzp_gamma = Fs + (delta_Wzp.transpose()*gWs).trace() +
+				(delta_Czp.transpose()*gCs).trace() +
+				(gamma/2)*nrm_delta_Wzp +
+				(gamma/2)*nrm_delta_Czp;
+
+			// break if delta is getting too small
+			if (r_sum <= 1e-20)
+			{
+				gradient_break = true;
+				break;
+			}
+
+			// break if objective at line searc point is smaller than Fzp_gamma
+			if (Fzp <= Fzp_gamma)
+				break;
+			else
+				gamma *= gamma_inc;
+		}
+
+		if (gradient_break)
+			break;
+
+		Wz_old = Wz;
+		Cz_old = Cz;
+		Wz = Wzp;
+		Cz = Czp;
+
+		// compute objective value
+		obj = Fzp;
+		for (task=0; task<n_tasks; task++)
+			obj += rho1*(Wz.col(task).norm());
+
+		// check if process should be terminated 
+		switch (options.termination)
+		{
+			case 0:
+				if (iter>=2)
+				{
+					if ( (CMath::abs(obj)-CMath::abs(obj_old)) <= options.tolerance)
+						break;
+				}
+			break;
+			case 1:
+				if (iter>=2)
+				{
+					if ( (CMath::abs(obj)-CMath::abs(obj_old)) <= options.tolerance*CMath::abs(obj_old))
+						break;
+				}
+			break;
+			case 2:
+				if (CMath::abs(obj) <= options.tolerance)
+					break;
+			break;
+			case 3:
+				if (iter>=options.max_iter)
+					break;
+			break;
+		}
+
+		iter++;
+		t_old = t;
+		t = 0.5 * (1 + CMath::sqrt(1.0 + 4*t*t));
+	}
+	SG_SDEBUG("%d iteration passed\n",iter);
+
+	SGMatrix<float64_t> tasks_w(n_feats, n_tasks);
+	for (int i=0; i<n_feats; i++)
+	{
+		for (int task=0; task<n_tasks; task++)
+			tasks_w[i] = Wzp(i,task);
+	}
+	SGVector<float64_t> tasks_c(n_tasks);
+	for (int i=0; i<n_tasks; i++) tasks_c[i] = Czp[i];
+	return slep_result_t(tasks_w, tasks_c);
+};
+};
diff --git a/src/shogun/lib/slep/malsar_joint_feature_learning.h b/src/shogun/lib/slep/malsar_joint_feature_learning.h
@@ -0,0 +1,28 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * Written (W) 2012 Sergey Lisitsyn
+ * Copyright (C) 2012 Jiayu Zhou and Jieping Ye
+ */
+
+#ifndef  MALSAR_JOINT_FEATURE_LEARNING_H_
+#define  MALSAR_JOINT_FEATURE_LEARNING_H_
+#include <shogun/lib/slep/slep_options.h>
+#include <shogun/features/DotFeatures.h>
+
+namespace shogun 
+{
+
+slep_result_t malsar_joint_feature_learning(
+		CDotFeatures* features,
+		double* y,
+		double rho1,
+		double rho2,
+		const slep_options& options);
+
+};
+#endif   /* ----- #ifndef MALSAR_JOINT_FEATURE_LEARNING_H_  ----- */
+
diff --git a/src/shogun/transfer/multitask/MultitaskL1L2LogisticRegression.cpp b/src/shogun/transfer/multitask/MultitaskL1L2LogisticRegression.cpp
@@ -0,0 +1,78 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * Copyright (C) 2012 Sergey Lisitsyn
+ */
+
+#include <shogun/transfer/multitask/MultitaskL1L2LogisticRegression.h>
+#include <shogun/lib/slep/malsar_joint_feature_learning.h>
+#include <shogun/lib/slep/slep_options.h>
+#include <shogun/lib/IndexBlockGroup.h>
+#include <shogun/lib/SGVector.h>
+
+namespace shogun
+{
+
+CMultitaskL1L2LogisticRegression::CMultitaskL1L2LogisticRegression() :
+	CMultitaskLogisticRegression(), m_rho1(0.0), m_rho2(0.0)
+{
+}
+
+CMultitaskL1L2LogisticRegression::CMultitaskL1L2LogisticRegression(
+     float64_t rho1, float64_t rho2, CDotFeatures* train_features, 
+     CBinaryLabels* train_labels, CIndexBlockRelation* task_relation) :
+	CMultitaskLogisticRegression(0.0,train_features,train_labels,task_relation)
+{
+	set_rho1(rho1);
+	set_rho2(rho2);
+}
+
+void CMultitaskL1L2LogisticRegression::set_rho1(float64_t rho1)
+{
+	m_rho1 = rho1;
+}
+
+void CMultitaskL1L2LogisticRegression::set_rho2(float64_t rho2)
+{
+	m_rho2 = rho2;
+}
+
+CMultitaskL1L2LogisticRegression::~CMultitaskL1L2LogisticRegression()
+{
+}
+
+bool CMultitaskL1L2LogisticRegression::train_machine(CFeatures* data)
+{
+	if (data && (CDotFeatures*)data)
+		set_features((CDotFeatures*)data);
+
+	ASSERT(features);
+	ASSERT(m_labels);
+
+	SGVector<float64_t> y(m_labels->get_num_labels());
+	for (int32_t i=0; i<y.vlen; i++)
+		y[i] = ((CBinaryLabels*)m_labels)->get_label(i);
+
+	slep_options options = slep_options::default_options();
+	options.termination = m_termination;
+	options.tolerance = m_tolerance;
+	options.max_iter = m_max_iter;
+	SGVector<index_t> ind = ((CIndexBlockGroup*)m_task_relation)->get_SLEP_ind();
+	options.ind = ind.vector;
+	options.n_tasks = ind.vlen-1;
+
+	slep_result_t model = malsar_joint_feature_learning(
+		features, y.vector, m_rho1, m_rho2, options);
+
+	m_tasks_w = model.w;
+	m_tasks_c = model.c;
+
+	ASSERT(m_task_relation);
+
+	return true;
+}
+
+}