This repository contains an implementation of a Multi-Layer Perceptron (MLP) from scratch using Python. It demonstrates the fundamental concepts of building and training a neural network, including forward propagation, backward propagation, and parameter optimization.

• Custom implementation of MLP without external machine learning libraries.
• Implements activation functions such as Tanh and ReLU.
• Supports gradient computation and backpropagation for weight updates.
• Includes training, validation, and testing phases with early stopping.
• Data preprocessing with normalization and handling missing values.

• value Class: Represents a scalar value with support for auto-differentiation.
• Neuron and Layers: Define the architecture of the MLP.
• Training Function: Trains the model with batch updates and monitors loss.

• Generates plots for:
  • Training Loss
  • Validation Loss
  • Testing Loss

• Python 3.x
• Required libraries: numpy, matplotlib, scipy, graphviz

1. Clone the repository:

   git clone https://github.com/your-username/MLP_from_Scratch.git
   

   cd MLP_from_Scratch

pip install -r requirements.txt

• Prepare the training, validation, and testing datasets.
• Configure the Model

• n_net = mlp(input_dim, [hidden_layers, ..., output_dim])

• Train the Model

• trained_net, train_loss, val_loss, test_loss = training(x_train, y_train, x_test, y_test, x_val, y_val, n_net, epochs, learning_rate)

• Visulaize the Loss

• plt.plot(train_loss)
  plt.title("Training Loss")
  plt.show()