AI-NeuralNet-Simple
view release on metacpan or search on metacpan
Revision historyforPerl extension AI::NeuralNet::Simple.0.11 November 18, 2006Converted from Inline::C to XSNo longerrequire5.008. 5.005 and above should be fine.0.10 December 29, 2005The following changes are all courtesy of Raphael Manfredi<Raphael_Manfredi [at] pobox.com>.Added tanh (bipolar) activation function.train_set() can nowacceptan error target to avoid over-training.Multiple network support.Persistence via storable.0.02 September 21 2005Added pod and pod coverage testsAdded Sub::Uplevel dependency to stop that annoying error failure :(0.01 Sat Jan 31 12:19:00 2004Applied patch from"Daniel L. Ashbrook"<anjiro [at] cc.gatech.edu>to fix a small memory allocation bug in infer()
double sigmoid(NEURAL_NETWORK*n, double val);double sigmoid_derivative(NEURAL_NETWORK*n, double val);float get_float_element(AV* array,intindex);intis_array_ref(SV*ref);void c_assign_random_weights(NEURAL_NETWORK *);void c_back_propagate(NEURAL_NETWORK *);void c_destroy_network(int);void c_feed(NEURAL_NETWORK *, double*input, double*output,intlearn);void c_feed_forward(NEURAL_NETWORK *);float c_get_learn_rate(int);void c_set_learn_rate(int, float);SV* c_export_network(inthandle);intc_import_network(SV *);#define ABS(x) ((x) > 0.0 ? (x) : -(x))intis_array_ref(SV*ref){if(SvROK(ref) && SvTYPE(SvRV(ref)) == SVt_PVAV)return1;else
returnhandle;}float c_get_learn_rate(inthandle){NEURAL_NETWORK*n= c_get_network(handle);returnn->learn_rate;}void c_set_learn_rate(inthandle, float rate){NEURAL_NETWORK*n= c_get_network(handle);n->learn_rate = rate;}double c_get_delta(inthandle){NEURAL_NETWORK*n= c_get_network(handle);returnn->delta;}void c_set_delta(inthandle, double delta){NEURAL_NETWORK*n= c_get_network(handle);n->delta = delta;}intc_get_use_bipolar(inthandle){NEURAL_NETWORK*n= c_get_network(handle);returnn->use_bipolar;}void c_set_use_bipolar(inthandle,intbipolar){NEURAL_NETWORK*n= c_get_network(handle);n->use_bipolar = bipolar;}intc_create_network(NEURAL_NETWORK*n){inti;/*eachof thenexttwo variableshasan extra rowforthe"bias"*/
if(!c_create_network(n))return-1;/* Perl already seeded the random number generator, via arand(1) call */c_assign_random_weights(n);returnhandle;}double c_train_set(inthandle, SV* set,intiterations, double mse){NEURAL_NETWORK*n= c_get_network(handle);AV*input_array,*output_array; /* perl arrays */double*input,*output; /* C arrays */double max_error = 0.0;intset_length=0;inti,j;intindex;set_length = av_len(get_array(set))+1;if(!set_length)croak("_train_set() array ref has no data");if(set_length % 2)croak("_train_set array ref must have an even number of elements");/* allocate memoryforout input and output arrays */input_array = get_array_from_aoa(set, 0);input = malloc(sizeof(double) * set_length * (av_len(input_array)+1));output_array = get_array_from_aoa(set, 1);output = malloc(sizeof(double) * set_length * (av_len(output_array)+1));for(i=0; i < set_length; i += 2) {input_array = get_array_from_aoa(set, i);if(av_len(input_array)+1 != n->size.input)croak("Length of input data does not match");/* iterate over the input_array and assign the floats to input */for(j = 0; j < n->size.input; j++) {index= (i/2*n->size.input)+j;input[index] = get_float_element(input_array, j);}output_array = get_array_from_aoa(set, i+1);if(av_len(output_array)+1 != n->size.output)croak("Length of output data does not match");for(j = 0; j < n->size.output; j++) {index= (i/2*n->size.output)+j;output[index] = get_float_element(output_array, j);}}for(i = 0; i < iterations; i++) {max_error = 0.0;for(j = 0; j < (set_length/2); j++) {double error;c_feed(n,&input[j*n->size.input],&output[j*n->size.output], 1);if(mse >= 0.0 || i == iterations - 1) {error = mean_square_error(n,&output[j*n->size.output]);if(error > max_error)max_error = error;}}
AV *get_array_from_aoa (aref,index)SV * arefintindexfloatc_get_learn_rate (handle)inthandlevoidc_set_learn_rate (handle, rate)inthandlefloat ratePREINIT:I32* temp;PPCODE:temp = PL_markstack_ptr++;c_set_learn_rate(handle, rate);if(PL_markstack_ptr != temp) {/* truly void, because dXSARGS not invoked */PL_markstack_ptr = temp;XSRETURN_EMPTY; /*returnempty stack */}/* must have used dXSARGS; list context implied */return; /* assume stack size is correct */doublec_get_delta (handle)inthandlevoidc_set_delta (handle, delta)inthandledouble deltaPREINIT:I32* temp;PPCODE:temp = PL_markstack_ptr++;c_set_delta(handle, delta);if(PL_markstack_ptr != temp) {/* truly void, because dXSARGS not invoked */PL_markstack_ptr = temp;XSRETURN_EMPTY; /*returnempty stack */}/* must have used dXSARGS; list context implied */return; /* assume stack size is correct */intc_get_use_bipolar (handle)inthandlevoidc_set_use_bipolar (handle, bipolar)inthandleintbipolarPREINIT:I32* temp;PPCODE:temp = PL_markstack_ptr++;c_set_use_bipolar(handle, bipolar);if(PL_markstack_ptr != temp) {/* truly void, because dXSARGS not invoked */PL_markstack_ptr = temp;XSRETURN_EMPTY; /*returnempty stack */}/* must have used dXSARGS; list context implied */return; /* assume stack size is correct */voidc_destroy_network (handle)
SV * inputSV * outputintc_new_network (input, hidden, output)intinputinthiddenintoutputdoublec_train_set (handle, set, iterations, mse)inthandleSV * setintiterationsdouble mseSV *c_infer (handle, array_ref)inthandleSV * array_ref
examples/game_ai.pl view on Meta::CPAN
my$net= AI::NeuralNet::Simple->new(4,20,4);$net->iterations(shift|| 100000);$net->train_set( [# health knife gun enemy[GOOD, YES, YES, 0], WANDER,[GOOD, YES, NO, 2], HIDE,[GOOD, YES, NO, 1], ATTACK,[GOOD, YES, NO, 0], WANDER,[GOOD, NO, YES, 2], ATTACK,[GOOD, NO, YES, 1], ATTACK,[GOOD, NO, NO, 3], HIDE,[GOOD, NO, NO, 2], HIDE,[GOOD, NO, NO, 1], RUN,
lib/AI/NeuralNet/Simple.pm view on Meta::CPAN
}my$seed=rand(1);# Perl invokes srand() on first call to rand()my$handle= c_new_network(@args);logdie"could not create new network"unless$handle>= 0;my$self=bless{input=>$args[0],hidden=>$args[1],output=>$args[2],handle=>$handle,},$class;$self->iterations(10000);# set a reasonable default}subtrain {my($self,$inputref,$outputref) =@_;returnc_train($self->handle,$inputref,$outputref);}subtrain_set {my($self,$set,$iterations,$mse) =@_;$iterations||=$self->iterations;$mse= -1.0unlessdefined$mse;returnc_train_set($self->handle,$set,$iterations,$mse);}subiterations {my($self,$iterations) =@_;if(defined$iterations) {logdie"iterations() value must be a positive integer."unless$iterationsand$iterations=~ /^\d+$/;$self->{iterations} =$iterations;return$self;}$self->{iterations};}subdelta {my($self,$delta) =@_;returnc_get_delta($self->handle )unlessdefined$delta;logdie"delta() value must be a positive number"unless$delta> 0.0;c_set_delta($self->handle,$delta);return$self;}subuse_bipolar {my($self,$bipolar) =@_;returnc_get_use_bipolar($self->handle )unlessdefined$bipolar;c_set_use_bipolar($self->handle,$bipolar);return$self;}subinfer {my($self,$data) =@_;c_infer($self->handle,$data);}subwinner {
lib/AI/NeuralNet/Simple.pm view on Meta::CPAN
$largest=$_if$arrayref->[$_] >$arrayref->[$largest];}return$largest;}sublearn_rate {my($self,$rate) =@_;returnc_get_learn_rate($self->handle )unlessdefined$rate;logdie"learn rate must be between 0 and 1, exclusive"unless$rate> 0 &&$rate< 1;c_set_learn_rate($self->handle,$rate);return$self;}subDESTROY {my$self=shift;c_destroy_network($self->handle );}## Serializing hook for Storable
lib/AI/NeuralNet/Simple.pm view on Meta::CPAN
just searchingwithGoogle.)Once the activation function is applied, the output is then sent through thenextsynapse, where it will be multiplied by w4 and the process willcontinue.=head2 C<AI::NeuralNet::Simple> architectureThe architecture used by this module has (at present) 3 fixed layers ofneurons: an input, hidden, and output layer. In practice, a 3 layer network isapplicable to many problems for which a neural network is appropriate, but thisis not always the case. In this module, we've settled on a fixed 3 layernetwork for simplicity.Here's how a three layer network might learn "logical or". First, we need todetermine how many inputs and outputs we'll have. The inputs are simple, we'llchoose two inputs as this is the minimum necessary to teach a network thisconcept. For the outputs, we'll also choose two neurons, with the neuron withthe highest output value being the "true" or "false" response that we arelooking for. We'll only have one neuron for the hidden layer. Thus, we get anetwork that resembles the following:
lib/AI/NeuralNet/Simple.pm view on Meta::CPAN
This is choosing the number of layers and the number of neurons per layer. InC<AI::NeuralNet::Simple>, the number of layers is fixed.With more complete neural net packages, you can also pick which activation=item 2 TrainingThis involves feeding the neural network enough data until the error rate islow enough to be acceptable. Often we have a large data set and merely keepiterating until the desired error rate is achieved.=item 3 Measuring resultsOne frequent mistake made with neural networks is failing to test the networkwith different data from the training data. It's quite possible for abackpropagation network to hit what is known as a "local minimum" which is nottruly where it should be. This will cause false results. To check for this,after training we often feed in other known good data for verification. If theresults are not satisfactory, perhaps a different number of neurons per layershould be tried or a different set of training data should be supplied.=back=head1 Programming C<AI::NeuralNet::Simple>=head2 C<new($input, $hidden, $output)>C<new()> accepts three integers. These number represent the number of nodes inthe input, hidden, and output layers, respectively. To create the "logical or"network described earlier:
lib/AI/NeuralNet/Simple.pm view on Meta::CPAN
T(x) = tanh(delta * x)tanh(x) = (exp(x) -exp(-x)) / (exp(x) +exp(-x))which allows the network to have neurons negatively impacting the weight,since T() is a signed function between (-1,+1) whereas S() only fallswithin (0,1).=head2 C<delta($delta)>Fetches the current I<delta> used in activation functions to scale thesignal, or sets the new I<delta>. The higher the delta, the steeper theactivation function will be. The argument must be strictly positive.You should not change I<delta> during the traning.=head2 C<use_bipolar($boolean)>Returns whether the network currently uses a bipolar activation function.If an argument is supplied, instruct the network to use a bipolar activationfunction or not.You should not change the activation function during the traning.=head2 C<train(\@input, \@output)>This method trains the network to associate the input data set with the outputdata set. Representing the "logical or" is as follows:$net->train([1,1] => [0,1]);$net->train([1,0] => [0,1]);$net->train([0,1] => [0,1]);$net->train([0,0] => [1,0]);Note that a one pass through the data is seldom sufficient to train a network.In the example "logical or" program, we actually run this data through thenetwork ten thousand times.for (1 .. 10000) {$net->train([1,1] => [0,1]);$net->train([1,0] => [0,1]);$net->train([0,1] => [0,1]);$net->train([0,0] => [1,0]);}The routine returns the Mean Squared Error (MSE) representing how far thenetwork answered.It is far preferable to use C<train_set()> as this lets you control the MSEover the training set and it is more efficient because there are less memorycopies back and forth.=head2 C<train_set(\@dataset, [$iterations, $mse])>Similar to train, this method allows us to train an entire data set at once.It is typically faster than calling individual "train" methods. The firstargument is expected to be an array ref of pairs of input and output arrayrefs.The second argument is the number of iterations to train the set. Ifthis argument is not provided here, you may use the C<iterations()> method toset it (prior to calling C<train_set()>, of course). A default of 10,000 willbe provided if not set.The third argument is the targeted Mean Square Error (MSE). When provided,the traning sequence will compute the maximum MSE seen during an iterationover the training set, and if it is less than the supplied target, thetraining stops. Computing the MSE at each iteration costs, but you arecertain to not over-train your network.$net->train_set([[1,1] => [0,1],[1,0] => [0,1],[0,1] => [0,1],[0,0] => [1,0],], 10000, 0.01);The routine returns the MSE of the last iteration, which is the highest MSEseen over the whole training set (and not an average MSE).=head2 C<iterations([$integer])>If called with a positive integer argument, this method will allow you to setnumber of iterations that train_set will use and will return the networkobject. If called without an argument, it will return the number of iterationsit was set to.$net->iterations; # returns 100000my @training_data = ([1,1] => [0,1],[1,0] => [0,1],[0,1] => [0,1],[0,0] => [1,0],);$net->iterations(100000) # let's have lots more iterations!->train_set(\@training_data);=head2 C<learn_rate($rate)>)This method, if called without an argument, will return the current learningrate. .20 is the default learning rate.If called with an argument, this argument must be greater than zero and lessthan one. This will set the learning rate and return the object.$net->learn_rate; #returns the learning rate$net->learn_rate(.1)->iterations(100000)->train_set(\@training_data);If you choose a lower learning rate, you will train the network slower, but youmay get a better accuracy. A higher learning rate will train the networkfaster, but it can have a tendancy to "overshoot" the answer when learning andnot learn as accurately.=head2 C<infer(\@input)>This method, if provided with an input array reference, will return an arrayreference corresponding to the output values that it is guessing. Note that
lib/AI/NeuralNet/Simple.pm view on Meta::CPAN
"AI Application Programming by M. Tim Jones, copyright (c) by Charles RiverMedia, Inc.The C code in this module is based heavily upon Mr. Jones backpropogationnetwork in the book. The"game ai"example in the examples directory is basedbecause it's more fun than many of the dry examples out there :)"Naturally Intelligent Systems", by Maureen Caudill and Charles Butler,copyright (c) 1990 by Massachussetts Institute of Technology.This book is a decent introduction to neural networks in general. The forwardfeed back error propogation is but one of many types.=head1 AUTHORSCurtis "Ovid" Poe, C<ovid [at] cpan [dot] org>Multiple network support, persistence, export of MSE (mean squared error),training until MSE below a given threshold and customization of the
t/10nn_simple.t view on Meta::CPAN
qr/^\QArguments to new() must be positive integers\E/,'... and supplying new() with bad arguments should also die';my$net=$CLASS->new(2,1,2);ok($net,'Calling new with good arguments should succeed');isa_ok($net,$CLASS=>'...and the object it returns');can_ok($net,'learn_rate');throws_ok {$net->learn_rate(2)}qr/^\QLearn rate must be between 0 and 1, exclusive\E/,'... and setting it outside of legal boundaries should die';is(sprintf("%.1f",$net->learn_rate),"0.2",'... and it should have the correct learn rate');isa_ok($net->learn_rate(.3),$CLASS=>'... and setting it should return the object');is(sprintf("%.1f",$net->learn_rate),"0.3",'... and should set it correctly');$net->learn_rate(.2);can_ok($net,'train');# teach the network logical 'or'ok($net->train([1,1], [0,1]),'Calling train() with valid data should succeed');for(1 .. 10000) {$net->train([1,1],[0,1]);$net->train([1,0],[0,1]);
t/10nn_simple.t view on Meta::CPAN
can_ok($net,'winner');is($net->winner([1,1]), 1,'... and it should return the index of the highest valued result');is($net->winner([1,0]), 1,'... and it should return the index of the highest valued result');is($net->winner([0,1]), 1,'... and it should return the index of the highest valued result');is($net->winner([0,0]), 0,'... and it should return the index of the highest valued result');# teach the network logical 'and' using the tanh() activation with delta=2$net=$CLASS->new(2,1,2);$net->delta(2);$net->use_bipolar(1);my$mse=$net->train_set([[1,1] => [0,1],[1,0] => [1,0],[0,1] => [1,0],[0,0] => [1,0],], 10000, 0.2);is($net->winner([1,1]), 1,'1 AND 1 = 1');is($net->winner([1,0]), 0,'1 AND 0 = 0');is($net->winner([0,1]), 0,'0 AND 1 = 0');is($net->winner([0,0]), 0,'0 AND 0 = 0');
t/20nn_multi.t view on Meta::CPAN
};can_ok($CLASS,'new');my$net1=$CLASS->new(2,1,2);ok($net1,'Calling new with good arguments should succeed');isa_ok($net1,$CLASS=>'...and the object it returns');can_ok($net1,'learn_rate');is(sprintf("%.1f",$net1->learn_rate),"0.2",'... and it should have the correct learn rate');isa_ok($net1->learn_rate(.5),$CLASS=>'... and setting it should return the object');is(sprintf("%.1f",$net1->learn_rate),"0.5",'... and should set it correctly');my$net2=$CLASS->new(5,8,2);ok($net2,'Calling new with good arguments should succeed');isa_ok($net2,$CLASS=>'...and the object it returns');can_ok($net2,'learn_rate');is(sprintf("%.1f",$net2->learn_rate),"0.2",'... and it should have the correct learn rate');isa_ok($net2->learn_rate(.3),$CLASS=>'... and setting it should return the object');is(sprintf("%.1f",$net2->learn_rate),"0.3",'... and should set it correctly');$net2->learn_rate(.2);
t/pod-coverage.t view on Meta::CPAN
build_rvc_destroy_networkc_export_networkc_get_deltac_get_learn_ratec_get_use_bipolarc_import_networkc_inferc_load_axac_new_networkc_set_deltac_set_learn_ratec_set_use_bipolarc_trainc_train_setget_arrayget_array_from_aoaget_elementget_float_elementhandleis_array_ref);pod_coverage_ok("AI::NeuralNet::Simple", {trustme=> [$ignore] } );
( run in 1.653 second using v1.01-cache-2.11-cpan-87723dcf8b7 )