https://github.com/tinsaeabreham45/rnn_numpy

Building Recurrent Neural Network Only Using Numpy
https://github.com/tinsaeabreham45/rnn_numpy

machine-learning nlp nlp-machine-learning numpy rnn rnn-model

Last synced: 10 days ago
JSON representation

Building Recurrent Neural Network Only Using Numpy

• Host: GitHub
• URL: https://github.com/tinsaeabreham45/rnn_numpy
• Owner: tinsaeabreham45
• Created: 2025-02-03T16:08:19.000Z (11 days ago)
• Default Branch: main
• Last Pushed: 2025-02-03T16:19:18.000Z (11 days ago)
• Last Synced: 2025-02-03T17:26:46.898Z (11 days ago)
• Topics: machine-learning, nlp, nlp-machine-learning, numpy, rnn, rnn-model
• Language: Jupyter Notebook
• Homepage:
• Size: 212 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# First RNN Implementation Using NumPy

This repository contains an implementation of a simple Recurrent Neural Network (RNN) using only NumPy. The goal is to understand the fundamental concepts of RNNs by manually implementing the forward propagation of an RNN cell and a full unrolled RNN across multiple time steps.

## Features

- **`rnn_cell_forward(xt, a_prev, parameters)`**: Computes the forward pass for a single RNN cell.

- **`rnn_forward(x, a0, parameters)`**: Implements the full forward propagation through time for an RNN.

## Implementation Details

- Uses **only NumPy** (no TensorFlow or PyTorch).

- Supports multiple time steps and batch processing.

- Applies the **tanh** activation function for hidden states.

- Outputs predictions using the **softmax** function.

## Purpose

This notebook serves as a foundational step toward understanding how RNNs work internally before moving on to more advanced deep learning frameworks like TensorFlow or PyTorch.