Handwritten-Digit-Recognition-Training-and-Evaluating-a-Neural-Network-numpy-no-TF-Keras-: A Jupyter Notebook repository from Navedshk

"Handwritten Digit Recognition: Training and Evaluating a Neural Network (numpy, no TF/Keras) "

README.md

# Digit Recognition Neural Network

This project implements a neural network to recognize handwritten digits from the MNIST dataset. The model is trained using a basic neural network architecture with one hidden layer and evaluated on both training and development sets. 

## Project Overview

The goal of this project is to build a neural network that can accurately classify images of handwritten digits (0-9). The implementation includes data preprocessing, forward and backward propagation, and evaluation of the model's performance.

## Files and Functions

- `init_params()`: Initializes the weights and biases of the neural network.
- `ReLU(Z)`: Applies the ReLU activation function.
- `softmax(Z)`: Applies the softmax activation function to get probabilities.
- `forward_prop(W1, b1, W2, b2, X)`: Performs forward propagation to compute activations.
- `ReLU_deriv(Z)`: Computes the derivative of the ReLU function.
- `one_hot(Y)`: Converts labels to one-hot encoded format.
- `backward_prop(Z1, A1, Z2, A2, W1, W2, X, Y)`: Computes gradients using backward propagation.
- `update_params(W1, b1, W2, b2, dW1, db1, dW2, db2, alpha)`: Updates weights and biases.
- `get_predictions(A2)`: Retrieves the predicted class from the output.
- `get_accuracy(predictions, Y)`: Computes the accuracy of predictions.
- `gradient_descent(X, Y, alpha, iterations)`: Trains the model using gradient descent.
- `make_predictions(X, W1, b1, W2, b2)`: Makes predictions for a given input.
- `test_prediction(index, W1, b1, W2, b2)`: Tests the model's prediction on a specific index and displays the result.

## Setup

1. **Install Dependencies**

   Make sure you have the following Python packages installed:
   - `numpy`
   - `pandas`
   - `matplotlib`

   You can install them using pip:
   ```bash
   pip install numpy pandas matplotlib

Download Dataset

Ensure the MNIST dataset is available in CSV format and located at /kaggle/input/digit-recognizer/train.csv.

Usage

Load Data Load the MNIST dataset using Pandas:

import pandas as pd
data = pd.read_csv('/kaggle/input/digit-recognizer/train.csv')

Preprocess Data Convert the data to a NumPy array, shuffle, and split into training and development sets.

Train the Model Initialize parameters and train the model:

W1, b1, W2, b2 = gradient_descent(X_train, Y_train, 0.10, 500)

Evaluate the Model Make predictions and evaluate accuracy:

dev_predictions = make_predictions(X_dev, W1, b1, W2, b2)
print(get_accuracy(dev_predictions, Y_dev))

Test Predictions Test and visualize predictions for specific images:
```
test_prediction(0, W1, b1, W2, b2)
```

Contact

For any questions or inquiries, please contact Naved Shaikh at Navedshaikh77920@gmail.com.

Happy coding!


Feel free to adjust any sections or add specific details about your project, such as example results or additional setup instructions.

Navedshk/Handwritten-Digit-Recognition-Training-and-Evaluating-a-Neural-Network-numpy-no-TF-Keras-

Usage

Contact