{"id":24634449,"url":"https://github.com/susanketsarkar/nn-from-scratch","last_synced_at":"2026-05-10T16:05:34.047Z","repository":{"id":273107784,"uuid":"918727271","full_name":"SusanketSarkar/NN-from-scratch","owner":"SusanketSarkar","description":"A pure NumPy implementation of a deep neural network, built for educational purposes and deep learning understanding. This project implements a multi-layer neural network with various features commonly found in modern deep learning frameworks.","archived":false,"fork":false,"pushed_at":"2025-01-18T17:52:37.000Z","size":23,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-03-20T07:47:17.452Z","etag":null,"topics":["deep-learning","implementation-of-algorithms","neural-network","numpy","python"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/SusanketSarkar.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2025-01-18T17:44:07.000Z","updated_at":"2025-01-18T17:52:38.000Z","dependencies_parsed_at":"2025-01-18T18:45:32.536Z","dependency_job_id":"3bfbacfa-f6e9-4294-80d5-9a5e98bb1854","html_url":"https://github.com/SusanketSarkar/NN-from-scratch","commit_stats":null,"previous_names":["susanketsarkar/nn-from-scratch"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/SusanketSarkar/NN-from-scratch","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SusanketSarkar%2FNN-from-scratch","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SusanketSarkar%2FNN-from-scratch/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SusanketSarkar%2FNN-from-scratch/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SusanketSarkar%2FNN-from-scratch/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/SusanketSarkar","download_url":"https://codeload.github.com/SusanketSarkar/NN-from-scratch/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SusanketSarkar%2FNN-from-scratch/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":268329555,"owners_count":24233064,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-08-02T02:00:12.353Z","response_time":74,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deep-learning","implementation-of-algorithms","neural-network","numpy","python"],"created_at":"2025-01-25T09:12:48.005Z","updated_at":"2026-05-10T16:05:33.949Z","avatar_url":"https://github.com/SusanketSarkar.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Neural Network Implementation from Scratch\n\nA pure NumPy implementation of a deep neural network, built for educational purposes and deep learning understanding. This project implements a multi-layer neural network with various features commonly found in modern deep learning frameworks.\n\n## Features\n\n- **Pure NumPy Implementation**: No deep learning frameworks used, everything built from scratch\n- **Modular Architecture**: Separate classes for layers, activations, optimizers, and loss functions\n- **Supported Components**:\n  - Dense (Fully Connected) Layers\n  - Dropout Layers for regularization\n  - ReLU and Sigmoid activation functions\n  - SGD optimizer with gradient clipping\n  - Binary Cross-entropy loss\n  - Various metrics calculations (Accuracy, Precision, Recall, F1-score)\n\n## Project Structure\n\n```\n├── neuralnet.py      # Main neural network implementation\n├── layers.py         # Layer implementations (Dense, Dropout)\n├── activation.py     # Activation functions\n├── optimizer.py      # Optimization algorithms\n├── loss.py          # Loss functions\n├── metrics.py       # Evaluation metrics\n└── implement.ipynb   # Example usage notebook\n```\n\n## Usage Example\n\n```python\nfrom neuralnet import NeuralNet\nfrom activation import ReLU, Sigmoid\nfrom optimizer import SGD\nfrom loss import Loss\n\n# Create network\nnn = NeuralNet()\nnn.dense(784, 128, ReLU())\nnn.dense(128, 64, ReLU())\nnn.dense(64, 10, Sigmoid())\n\n# Compile\nnn.compile(SGD(learning_rate=0.01), Loss.binary_crossentropy)\n\n# Train\nhistory = nn.fit(x_train, y_train, \n                epochs=10, \n                batch_size=32, \n                val_split=0.2, \n                verbose=2)\n\n# Predict\npredictions = nn.predict(x_test)\n```\n\n## Features in Detail\n\n### Layers\n- **Dense Layer**: Fully connected layer with configurable input/output dimensions\n- **Dropout Layer**: Regularization layer with configurable dropout rate\n\n### Activation Functions\n- **ReLU**: Rectified Linear Unit\n- **Sigmoid**: Sigmoid activation for binary classification\n\n### Optimizers\n- **SGD**: Stochastic Gradient Descent with gradient clipping\n\n### Loss Functions\n- **Binary Cross-entropy**: For classification tasks\n\n### Metrics\n- Accuracy\n- Precision\n- Recall\n- F1-score\n- Confusion Matrix\n\n## Requirements\n\n- NumPy\n- scikit-learn (for metrics and data splitting)\n- tqdm (for progress bars)\n- matplotlib (for visualization)\n\n## Installation\n\n```bash\ngit clone https://github.com/yourusername/neural-network-from-scratch.git\ncd neural-network-from-scratch\npip install -r requirements.txt\n```\n\n## Example Results\n\nThe network has been tested on the MNIST dataset, achieving reasonable performance for a from-scratch implementation:\n- Training accuracy: 99.445%\n- Validation accuracy: 97.625%\n- {'accuracy': 0.97625,\n 'precision': 0.9759900486865305,\n 'recall': 0.9764165113480248,\n 'f1_score': 0.9761682848191737}\n\n## Contributing\n\nFeel free to submit issues and enhancement requests!\n\n## License\n\nThis project is licensed under the MIT License - see the LICENSE file for details.\n\n## Acknowledgments\n\n- Built as an educational project to understand deep learning fundamentals\n- Inspired by modern deep learning frameworks like TensorFlow and PyTorch\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsusanketsarkar%2Fnn-from-scratch","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsusanketsarkar%2Fnn-from-scratch","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsusanketsarkar%2Fnn-from-scratch/lists"}