A unified mathematical and computational framework for non-Markovian adaptive systems with explicit multi-timescale memory. Built on Geometric Fourier Extension (GFE) and stabilized Sum-of-Exponentials (SOE) kernels, with rigorous stability proofs, spectral units, and hardware-mappable implementations. Julia reference and a C++ core in development

Сalculation of cryptographic characteristics for a cipher based on a generalized feistel network

This repository explores the classic XOR problem — a foundational challenge in neural network history — and demonstrates how a simple multi-layer perceptron (MLP) can solve it using non-linear activation functions.