Madrin
A cute Neural Network library with Keras-like API. 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](https://numpy.org/install/)
Installation
shell
pip install madrin
Demo
**Create a neural network:**<br>
You can create a Neural Network by passing a list of layers to the `Network` constructor.<br>
Currently it supports the following layers:
>`Linear(no_of_neurons, input_size)`<br>
>`Relu()`<br>
>`LeakyRelu()`<br>
>`Sigmoid()`<br>
>`Tanh()`<br>
>`Softmax()`<br>
python
import numpy as np
Import the necessary classes from the madrin library
from madrin import Linear, Sigmoid, Relu, LeakyRelu, Tanh, Softmax, Network
Generate some dummy data for training
np.random.seed(0) For reproducibility
X_train = np.random.randn(1000, 3) 1000 samples, 3 features each
y_train = np.random.randint(0, 3, 1000) 1000 labels (3 classes)
Create the network
model = Network([
Linear(no_of_neurons=5, input_size=3), First layer: 3 input features, 5 neurons
Relu(), ReLU activation
Linear(no_of_neurons=3, input_size=5), Second layer: 5 input features, 3 neurons (output layer)
Softmax() Softmax activation for multi-class classification
])
See the total number of trainable parameters(i.e., weights and biases)
print(model.n_parameters())
Compile the network with loss function and learning rate
model.compile(loss='categorical_crossentropy', lr=0.01)
Train the network
model.fit(X_train, y_train, epochs=1000, batch_size=100, track_loss = True)
Make predictions
predictions = model.forward(X_train)
Print the predictions
print(predictions)
Print the training costs
import matplotlib.pyplot as plt
plt.plot(np.arange(len(model.costs)),model.costs)
plt.xlabel('Epochs')
plt.ylabel('Cost')
plt.title('Training Cost Over Time')
plt.show()
Contributing
Contributions are welcome! Please open an issue or submit a pull request on [Github](https://github.com/manohar3000/Madrin-A_Neural_Network_Library).
License
Madrin is released under the MIT License.