#include <stdio.h>
#include <vector>
#include <string>
#include <math.h>
#include "mnist_common.h"
float accuracy(
const array& predicted,
const array& target)
{
array val, plabels, tlabels;
max(val, tlabels, target, 1);
max(val, plabels, predicted, 1);
return 100 * count<float>(plabels == tlabels) / tlabels.
elements();
}
float abserr(
const array& predicted,
const array& target)
{
return 100 * sum<float>(
abs(predicted - target)) / predicted.
elements();
}
{
}
const array &X,
const array &Y,
double lambda = 1.0)
{
array H = predict(X, Weights);
array Jreg = 0.5 *
sum(lambdat * Weights * Weights);
J = (Jerr + Jreg) / m;
dJ = (
matmulTN(X, D) + lambdat * Weights) / m;
}
double alpha = 0.1,
double lambda = 1.0,
double maxerr = 0.01,
int maxiter = 1000,
bool verbose = false)
{
float err = 0;
for (int i = 0; i < maxiter; i++) {
cost(J, dJ, Weights, X, Y, lambda);
err = max<float>(
abs(J));
if (err < maxerr) {
printf("Iteration %4d Err: %.4f\n", i + 1, err);
printf("Training converged\n");
return Weights;
}
if (verbose && ((i + 1) % 10 == 0)) {
printf("Iteration %4d Err: %.4f\n", i + 1, err);
}
Weights = Weights - alpha * dJ;
}
printf("Training stopped after %d iterations\n", maxiter);
return Weights;
}
void benchmark_logistic_regression(
const array &train_feats,
const array &train_targets,
{
timer::start();
array Weights = train(train_feats, train_targets, 0.1, 1.0, 0.01, 1000);
printf("Training time: %4.4lf s\n", timer::stop());
timer::start();
const int iter = 100;
for (int i = 0; i < iter; i++) {
array test_outputs = predict(test_feats , Weights);
}
printf("Prediction time: %4.4lf s\n", timer::stop() / iter);
}
int logit_demo(bool console, int perc)
{
array train_images, train_targets;
array test_images, test_targets;
int num_train, num_test, num_classes;
float frac = (float)(perc) / 100.0;
setup_mnist<true>(&num_classes, &num_train, &num_test,
train_images, test_images,
train_targets, test_targets, frac);
int feature_length = train_images.
elements() / num_train;
array train_feats =
moddims(train_images, feature_length, num_train).
T();
array test_feats =
moddims(test_images , feature_length, num_test ).
T();
train_targets = train_targets.
T();
test_targets = test_targets.
T();
train_feats =
join(1,
constant(1, num_train, 1), train_feats);
test_feats =
join(1,
constant(1, num_test , 1), test_feats );
array Weights = train(train_feats, train_targets,
0.1,
1.0,
0.01,
1000,
true);
array train_outputs = predict(train_feats, Weights);
array test_outputs = predict(test_feats , Weights);
printf("Accuracy on training data: %2.2f\n",
accuracy(train_outputs, train_targets ));
printf("Accuracy on testing data: %2.2f\n",
accuracy(test_outputs , test_targets ));
printf("Maximum error on testing data: %2.2f\n",
abserr(test_outputs , test_targets ));
benchmark_logistic_regression(train_feats, train_targets, test_feats);
if (!console) {
test_outputs = test_outputs.
T();
display_results<true>(test_images, test_outputs, test_targets.
T(), 20);
}
return 0;
}
int main(int argc, char** argv)
{
int device = argc > 1 ? atoi(argv[1]) : 0;
bool console = argc > 2 ? argv[2][0] == '-' : false;
int perc = argc > 3 ? atoi(argv[3]) : 60;
try {
return logit_demo(console, perc);
std::cerr << ae.
what() << std::endl;
}
return 0;
}
A multi dimensional data container.
Definition: array.h:27
Definition: exception.h:20
virtual const char * what() const
Definition: exception.h:34
AFAPI array abs(const array &in)
C++ Interface for absolute value.
AFAPI array log(const array &in)
C++ Interface for natural logarithm.
AFAPI array matmulTN(const array &lhs, const array &rhs)
Matrix multiply of two arrays.
AFAPI array matmul(const array &lhs, const array &rhs, const matProp optLhs=AF_MAT_NONE, const matProp optRhs=AF_MAT_NONE)
Matrix multiply of two arrays.
array constant(T val, const dim4 &dims, const dtype ty=(af_dtype) dtype_traits< T >::ctype)
AFAPI void setDevice(const int device)
Sets the current device.
AFAPI void sync(const int device=-1)
Blocks until the device is finished processing.
AFAPI array join(const int dim, const array &first, const array &second)
Join 2 arrays along dim.
AFAPI array moddims(const array &in, const unsigned ndims, const dim_t *const dims)
dim4 dims() const
Get dimensions of the array.
void eval() const
Evaluate any JIT expressions to generate data for the array.
array T() const
Get the transposed the array.
dim_t elements() const
get the number of elements in array
AFAPI array max(const array &in, const int dim=-1)
C++ Interface for maximum values in an array.
AFAPI array sum(const array &in, const int dim=-1)
C++ Interface for sum of elements in an array.
Definition: algorithm.h:15
AFAPI array sigmoid(const array &in)
C++ Interface for calculating sigmoid function of an array.