/MNIST_without_tensorflow

Primary LanguageJupyter Notebook

Digit Recognition with Neural Network (NumPy and Pandas)

Overview

This project implements a simple neural network for digit recognition using only NumPy and Pandas. The neural network architecture consists of three layers: two hidden layers with 16 nodes each and an output layer with 10 nodes corresponding to the digits 0-9.

Features

  • Neural Network Architecture:

    • Input Layer: 784 nodes (28x28 pixels for each digit image)
    • Hidden Layer 1: 16 nodes
    • Hidden Layer 2: 16 nodes
    • Output Layer: 10 nodes (0-9 digits)

  • Activation Function:

    • Rectified Linear Unit (ReLU) for hidden layers
    • Softmax for the output layer

  • Training Data:

    • MNIST dataset used for training and testing
    • The data used was the data provided for the Kaggle Digit Recognizer Task
    • I downloaded the training data and split it into training and testing sets.
    • 20% for the testing set and 80% for the training set.

  • Dependencies:

    • NumPy for numerical operations
    • Pandas for data manipulation

Results

  • Training Accuracy: [90%]
  • Testing Accuracy: [89%]

Acknowledgments

  • The MNIST dataset is used for training and testing, and it can be found here.

License

This project is licensed under the MIT License - see the LICENSE file for details.