https://github.com/akash-nath29/mlp_network_from_scratch
Multi Layer Perceptron from scratch able to detect Handwritten Digits
https://github.com/akash-nath29/mlp_network_from_scratch
Last synced: 2 months ago
JSON representation
Multi Layer Perceptron from scratch able to detect Handwritten Digits
- Host: GitHub
- URL: https://github.com/akash-nath29/mlp_network_from_scratch
- Owner: Akash-nath29
- Created: 2024-12-31T08:18:03.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2024-12-31T09:03:36.000Z (4 months ago)
- Last Synced: 2024-12-31T09:25:39.295Z (4 months ago)
- Language: Jupyter Notebook
- Size: 259 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# MLP Network from Scratch
This project implements a handwritten digit classifier using a Multi-Layer Perceptron (MLP) neural network. The model is trained on the MNIST dataset and can predict digits from 0 to 9.
## Project Structure
```
MLP_Network_from_scratch/
│
├── dataloader/
│ └── dataloader.py
├── functions/
│ └── functions.py
├── neuralnet/
│ └── model.py
├── data/
│ ├── mnist_train.csv
│ └── mnist_test.csv
├── train.py
├── requirements.txt
├── README.md
|── test.py
├── model.ipynb
└── .gitignore
```
## Setup Instructions
1. Clone the repository:
```sh
git clone https://github.com/Akash-nath29/MLP_Network_from_scratch.git
cd MLP_Network_from_scratch
```
2. Install the required dependencies:
```sh
pip install -r requirements.txt
```
3. Download the MNIST dataset and place the CSV files in the `data/` directory.
## Usage
### Training the Model
To train the model, run the `train.py` script:
```sh
python train.py
```
This will train the MLP neural network on the MNIST dataset and save the trained model to `model.json`.
### Testing the Model
To test the model, run the `test.py` script:
```sh
uvicorn test:app --reload
```
This will start a FastAPI server that uses the model to predict handwritten digits.
### Evaluating the Model
You can evaluate the model and visualize the predictions using the Jupyter notebook `model.ipynb`. Open the notebook and run the cells to see the results.
### Plotting Predictions
The `plot_predictions` function in `train.py` can be used to visualize the model's predictions on the test set.
## Functions and Classes
### Data Loader
- `load_data(train_csv, test_csv)`: Loads and normalizes the training and testing data from CSV files.
### Functions
- `one_hot_encoding(y, num_classes)`: Converts labels to one-hot encoded vectors.
- `sigmoid(z)`: Sigmoid activation function.
- `sigmoid_derivative(z)`: Derivative of the sigmoid function.
- `softmax(z)`: Softmax activation function.
- `cross_entropy_loss(y_true, y_pred)`: Computes the cross-entropy loss.
- `accuracy(y_true, y_pred)`: Computes the accuracy of predictions.
### Neural Network Model
- `MLPNetwork`: Class implementing the MLP neural network with methods for forward pass, backward pass, training, prediction, and loss plotting.
## License
This project is licensed under the [MIT LICENSE](LICENSE).