madrin 0.0.1

No project description provided

Madrin

A cute Neural Network library with Keras-like API with just 100 lines of code. Build for fun and educational purposes. Because the code is so simple, it is very easy to change to your needs. Still under active development.

Dependencies

• numpy

Demo

Create a neural network:
You can create a Neural Network by passing a list of layers to the Network constructor.
Currently it supports the following layers:

Linear(no_of_neurons, input_size)
Relu()
Sigmoid()
Tanh()
Softmax()

import madrin as mad

model = Network([ mad.Linear(50,1), mad.Relu(),
                  mad.Linear(50, 50), mad.Relu(),
                  mad.Linear(50, 50), mad.Relu(),
                  mad.Linear(1,50), mad.Sigmoid() ])

Training

model.fit(inputs, labels, learning_rate, epochs, loss='mse') 

Evaluate

model.costs(inputs, labels)

Predictions

model.forward(inputs)

Visualize the learning

import numpy as np
import matplotlib.pyplot as plt
plt.plot(np.arange(len(model.costs)),model.costs)
plt.xlabel("Iteration")
plt.ylabel("Loss")
plt.show()

Visualize the predictions(if possible)

plt.scatter(inputs, labels)
plt.plot(inputs, model.forward(inputs))
plt.show()