Show HN: Static-allocation MLP inference in ANSI C using a 2-slot ring buffer (github.com)

A new header-only library for implementing Multi-Layer Perceptrons (MLPs) in ANSI C has been announced, featuring a static allocation method that utilizes a 2-slot circular buffer for efficient feedforward predictions. This approach aims to optimize both speed and memory usage, making the library particularly useful for resource-constrained environments like AVR microcontrollers. Notably, the library supports int-quantized weights, making it compatible with Arduino systems, and is designed to be both portable and simple to use. The significance of this development lies in its ability to provide a lightweight and efficient alternative for deploying neural networks on small, embedded systems without sacrificing performance. The library's compatibility with quantization techniques allows developers to optimize MLP model accuracy while reducing the required memory footprint. This opens up new avenues for deploying AI models in edge computing scenarios, where computational resources are limited. With an example demonstrating the training of an MLP to emulate a double XOR gate, the library invites users to explore its straightforward setup and capabilities in practical applications.