PLearn 0.1
|
00001 // -*- C++ -*- 00002 00003 // DeepFeatureExtractorNNet.h 00004 // 00005 // Copyright (C) 2006 Hugo Larochelle 00006 // 00007 // Redistribution and use in source and binary forms, with or without 00008 // modification, are permitted provided that the following conditions are met: 00009 // 00010 // 1. Redistributions of source code must retain the above copyright 00011 // notice, this list of conditions and the following disclaimer. 00012 // 00013 // 2. Redistributions in binary form must reproduce the above copyright 00014 // notice, this list of conditions and the following disclaimer in the 00015 // documentation and/or other materials provided with the distribution. 00016 // 00017 // 3. The name of the authors may not be used to endorse or promote 00018 // products derived from this software without specific prior written 00019 // permission. 00020 // 00021 // THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR 00022 // IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 00023 // OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN 00024 // NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 00025 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED 00026 // TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 00027 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 00028 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 00029 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 00030 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 00031 // 00032 // This file is part of the PLearn library. For more information on the PLearn 00033 // library, go to the PLearn Web site at www.plearn.org 00034 00035 /* ******************************************************* 00036 * $Id: .pyskeleton_header 544 2003-09-01 00:05:31Z plearner $ 00037 ******************************************************* */ 00038 00039 // Authors: Hugo Larochelle 00040 00044 #ifndef DeepFeatureExtractorNNet_INC 00045 #define DeepFeatureExtractorNNet_INC 00046 00047 #include <plearn_learners/generic/PLearner.h> 00048 #include <plearn/vmat/AppendNeighborsVMatrix.h> 00049 #include <plearn/opt/Optimizer.h> 00050 #include <plearn/var/VarArray.h> 00051 00052 namespace PLearn { 00053 00061 class DeepFeatureExtractorNNet : public PLearner 00062 { 00063 typedef PLearner inherited; 00064 00065 public: 00066 //##### Public Build Options ############################################ 00067 00069 TVec<int> nhidden_schedule; 00071 PP<Optimizer> optimizer; 00075 PP<Optimizer> optimizer_supervised; 00077 int batch_size; 00079 int batch_size_supervised; 00081 string hidden_transfer_func; 00083 string output_transfer_func; 00085 int nhidden_schedule_position; 00087 TVec<string> cost_funcs; 00089 real weight_decay; 00091 real bias_decay; 00093 string penalty_type; 00096 real classification_regularizer; 00099 real regularizer; 00102 real margin; 00104 string initialization_method; 00106 Vec paramsvalues; 00108 int noutputs; 00111 bool use_same_input_and_output_weights; 00115 bool always_reconstruct_input; 00118 bool use_activations_with_cubed_input; 00120 int use_n_first_as_supervised; 00122 bool use_only_supervised_part; 00127 real relative_minimum_improvement; 00133 string input_reconstruction_error; 00138 real autoassociator_regularisation_weight; 00145 real supervised_signal_weight; 00147 int k_nearest_neighbors_reconstruction; 00148 00149 public: 00150 //##### Public Member Functions ######################################### 00151 00153 DeepFeatureExtractorNNet(); 00154 00155 00156 //##### PLearner Member Functions ####################################### 00157 00160 virtual int outputsize() const; 00161 00165 virtual void forget(); 00166 00170 virtual void train(); 00171 00173 virtual void computeOutput(const Vec& input, Vec& output) const; 00174 00176 virtual void computeCostsFromOutputs(const Vec& input, const Vec& output, 00177 const Vec& target, Vec& costs) const; 00178 00181 virtual TVec<std::string> getTestCostNames() const; 00182 00185 virtual TVec<std::string> getTrainCostNames() const; 00186 00187 00188 // *** SUBCLASS WRITING: *** 00189 // While in general not necessary, in case of particular needs 00190 // (efficiency concerns for ex) you may also want to overload 00191 // some of the following methods: 00192 virtual void computeOutputAndCosts(const Vec& input, const Vec& target, Vec& output, Vec& costs) const; 00193 // virtual void computeCostsOnly(const Vec& input, const Vec& target, Vec& costs) const; 00194 // virtual void test(VMat testset, PP<VecStatsCollector> test_stats, VMat testoutputs=0, VMat testcosts=0) const; 00195 // virtual int nTestCosts() const; 00196 // virtual int nTrainCosts() const; 00197 // virtual void resetInternalState(); 00198 // virtual bool isStatefulLearner() const; 00199 00200 00201 //##### PLearn::Object Protocol ######################################### 00202 00203 // Declares other standard object methods. 00204 PLEARN_DECLARE_OBJECT(DeepFeatureExtractorNNet); 00205 00206 // Simply calls inherited::build() then build_() 00207 virtual void build(); 00208 00210 virtual void makeDeepCopyFromShallowCopy(CopiesMap& copies); 00211 00212 protected: 00213 //##### Protected Options ############################################### 00214 00220 int nhidden_schedule_current_position; 00222 VarArray params; 00224 VarArray params_to_train; 00226 VarArray weights; 00228 VarArray reconstruction_weights; 00230 VarArray biases; 00232 VarArray invars; 00234 Var input; 00236 Var output; 00238 Var feature_vector; 00240 Var hidden_representation; 00242 Var neighbor_indices; 00244 Var target; 00246 Var unsupervised_target; 00248 Var sampleweight; 00250 VarArray costs; 00252 VarArray penalties; 00254 Var training_cost; 00256 Var test_costs; 00258 VMat sup_train_set; 00260 VMat unsup_train_set; 00262 PP<AppendNeighborsVMatrix> knn_train_set; 00263 00265 mutable Func f; 00267 mutable Func test_costf; 00269 mutable Func output_and_target_to_cost; 00271 mutable Func to_feature_vector; 00274 TVec<VarArray> autoassociator_params; 00277 VarArray autoassociator_training_costs; 00278 00279 protected: 00280 //##### Protected Member Functions ###################################### 00281 00283 static void declareOptions(OptionList& ol); 00284 00288 Var hiddenLayer(const Var& input, const Var& weights, string transfer_func, 00289 Var& before_transfer_function, bool use_cubed_value=false); 00290 00294 Var hiddenLayer(const Var& input, const Var& weights, const Var& bias, 00295 bool transpose_weights, string transfer_func, 00296 Var& before_transfer_function, bool use_cubed_value=false); 00297 00302 void buildOutputFromInput(const Var& the_input, Var& hidden_layer, 00303 Var& before_transfer_func); 00304 00306 void buildTargetAndWeight(); 00307 00309 void buildCosts(const Var& output, const Var& target, 00310 const Var& unsupervised_target, 00311 const Var& before_transfer_func, const Var& output_sup); 00312 00314 void buildFuncs(const Var& the_input, const Var& the_output, 00315 const Var& the_target, const Var& the_sampleweight); 00316 00321 void fillWeights(const Var& weights, bool fill_first_row, 00322 real fill_with_this=0); 00323 00325 virtual void buildPenalties(); 00326 00327 private: 00328 //##### Private Member Functions ######################################## 00329 00331 void build_(); 00332 00333 private: 00334 //##### Private Data Members ############################################ 00335 00336 // The rest of the private stuff goes here 00337 }; 00338 00339 // Declares a few other classes and functions related to this class 00340 DECLARE_OBJECT_PTR(DeepFeatureExtractorNNet); 00341 00342 } // end of namespace PLearn 00343 00344 #endif 00345 00346 00347 /* 00348 Local Variables: 00349 mode:c++ 00350 c-basic-offset:4 00351 c-file-style:"stroustrup" 00352 c-file-offsets:((innamespace . 0)(inline-open . 0)) 00353 indent-tabs-mode:nil 00354 fill-column:79 00355 End: 00356 */ 00357 // vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=8:softtabstop=4:encoding=utf-8:textwidth=79 :