Wolf — A Fast, Native C++ Neural Net Library

Wolf is a barebones C++23 neural network implementation. Trains a fully connected network at blazing fast speeds on CPU.

Documentation

• Wolf — A Fast, Native C++ Neural Net Library
• Documentation
  • Features
    • Build from Source (CMake)
  • External Libraries Used (No need to install)
    • References

---

Features

• Fully connected feed-forward neural networks
• Backpropagation + Stochastic Gradient Descent
• Linear and ReLU Layers
• Fully optimized for CPU
• Batched Learning
• Helper functions to save and load neural nets
• Adam, Momentum and RMSProp Optimizer
• MSE, Cross Entropy and Binary Cross Entropy cross

---

Build from Source (CMake)

Requirements: Recent C++ Compiler (gcc > 14)

1. Clone and build

git clone https://github.com/warg-void/Wolf.git
cd Wolf
cmake -B build -DBUILD_MNIST ON
cmake --build build

2. Run the examples

./build/examples/xortest
./build/examples/irisClassifier
./build/examples/mnistClassifier

External Libraries Used (No need to install)

• OpenMP (for parallelization)
• zpp_bits (for serializing and deserializing)

References

1. Deep Learning: Foundations and Concepts: by Christopher M. Bishop and Hugh Bishop, Springer Cham (2023).