Research indexJournal article
The OPi Transform Framework: Binary Complement Learning and Sequential Pattern Generation through Nonlinear Transform Kernels
Basta
Figshare1 Jan 2025Open access
NobleIDark:/48914/w1/20260525/E326B815DOI10.6084/m9.figshare.30370546.v1
The OPi Transform Framework: Binary Complement Learning and Sequential Pattern Generation through Nonlinear Transform Kernels
Basta · Figshare · 2025
NobleIDark:/48914/w1/20260525/E326B815DOI10.6084/m9.figshare.30370546.v1
Abstract
This work introduces and validates the *OPi Transform*, a nonlinear kernel layer designed to augment neural network architectures for both logical and sequential learning tasks. The OPi layer encodes nonlinearities inspired by trigonometric-logarithmic coupling, providing a rich feature transformation mechanism capable of generalizing across binary and text domains. We first demonstrate the OPi layer’s efficiency through a *binary complement learning task*, in which the model learns to predict the logical inverse of binary vectors. Scaling experiments across datasets of 1K, 10K, and 100K samples reveal near-perfect accuracy (99.71%) and decreasing loss, establishing the OPi layer’s stability and scalability. Extending beyond logical tasks, we embed the OPi Transform within a recurrent architecture (Embedding–OPi–LSTM–Dense) to construct a *text generation model*. The system effectively learns and reproduces stylistic sequences from philosophical texts, confirming the layer’s adaptability to sequential data. The OPi Transform thus provides a unifying kernel structure capable of abstract nonlinear mapping, generalizing from discrete logical operations to contextual linguistic understanding—demonstrating potential applications in explainable AI, symbolic reasoning, and neuromorphic computation