Metadata-Version: 2.1
Name: DirNet
Version: 1.1.1
Summary: Lightweight library for neural network-based directional prediction
Home-page: https://github.com/Foloip
Author: foloip
Author-email: NeuralNetworkTools@gmail.com
License: MIT
Keywords: neural network,direction prediction,ai,machine learning,Perceptron,Backpropagation,relu,sigmoid,loss
Platform: UNKNOWN
Description-Content-Type: text/markdown
License-File: LICENSE

You need to install numpy

Activation finction

ReLU: relu(x)
Relu derivative: relu_derivative(x)
Sigmoid: sigmoid(x)
Sigmoid derivative: sigmoid_derivative(x)
Sigmoid derivative (without activation): sigmoid_derivative_from_output(s)
Leaky ReLU: leaky_relu(x, alpha=0.01)
Leaky ReLU derivative: 
Soft max: softmax(z)

Loss

Cross entropy: cross_entropy(y_true, y_pred)
MSE Loss: mse_loss(y_true, y_pred)

Metrics

Accuracy: accuracy(output, tags)

Utils

Stats: statistics(loss, error, output, epoch, accuracy)


There is feature in version 1.1.1. Thank for download

Made by: foloip
Github: https://github.com/Foloip

