deeplearning-models is a collection of deep learning architectures and implementations in Jupyter Notebooks for TensorFlow and PyTorch.
The project addresses the need for reference implementations of foundational and modern deep learning models. It organizes notebooks by architecture type—traditional machine learning, multilayer perceptrons, convolutional neural networks, and others—providing working code for each approach. The collection spans from basic implementations like perceptron and logistic regression through to more complex architectures, with some notebooks including implementations built from scratch to illustrate underlying mechanics.
This collection suits developers and students learning deep learning who want to see concrete implementations across multiple frameworks. The dual-framework approach means you can compare how the same model is built in PyTorch versus TensorFlow, which is valuable for understanding framework-specific patterns. The inclusion of implementations from scratch, such as multilayer perceptron with backpropagation implemented manually, helps bridge the gap between theory and framework-based practice. The notebooks use standard datasets like MNIST and Iris, making results reproducible and comparable.
The project maintains implementations across PyTorch, PyTorch Lightning, and TensorFlow, indicating ongoing effort to keep pace with framework evolution. Notebooks are organized into clearly delineated sections covering progressively more sophisticated architectures, suggesting a structured approach to content curation. The repository includes both basic implementations and variations that explore specific techniques like dropout and batch normalization applied to the same base architecture, supporting comparative learning.
