Skip to content

Latest commit

 

History

History
34 lines (18 loc) · 1.98 KB

README.md

File metadata and controls

34 lines (18 loc) · 1.98 KB
Raw

LlamaNet

A custom built neural network library built from scratch in C++

Description:

Hi 👋! I'm Alex, and I developed a Neural Network deep learning library in C++ from scratch, using only the standard C++ libraries. The library features cutting-edge capabilities for building, training, and deploying neural networks and incorporates key advancements, such as the Adaptive Moment Estimation algorithm for adaptive learning rate adjustments and a fully customizable neural network architecture.

Details:

  • Custom Neural Network Architecture: Designed and implemented fully customizable layers and activation functions (ReLU, Sigmoid, Tanh, etc.), allowing for custom network topology and training parameters.

  • Optimization: Implemented Adam optimization for efficient and adaptive gradient descent, ensuring faster convergence, alongside Xavier initialization to improve weight initialization and promote stable gradient flow during training.

  • Efficient Training: Enabled randomized mini-batch gradient descent with advanced backpropagation, supporting loss/cost functions like Huber Loss.

  • Scalable Design: Built a modular framework supporting different topologies, making it easy to expand and adapt for various datasets and applications.

Testing:

  • Achieved 98.10% accuracy over 10,000 test cases on the MNIST Handwritten Digits Classification Test

  • Achieved 96.26% accuracy after a single epoch.

  • See the Excel spreadsheet for a more in-depth analysis of the library's test results :D

Future Changes:

  • Implement parallelization to improve computing time

  • Implement cross-entropy cost function and softmax to optimize classification tasks

  • Implement SVMs

You'll probably see these changes in v2 of the LlamaNet, where I will also implement a custom image processing library to be able to detect and classify hotdogs 🌭🌭🌭🌭. Hope to see you there!