Deep Neural Networks (DNNs) exhibit acute vulnerabilities to intermediate activation layer perturbations engineered through out-of-distribution (OOD) noise injection and feature-steering gradient updates. Conventional defensive paradigms—such as adversarial retraining or external pre-processing guardrails—introduce severe computational latency penalties (≥10 ms) and cause clean distribution performance degradation. This paper presents Project ASTRA (Zero-retraining Vector Interception Layer v2.0), a parameter-invariant runtime framework designed to dynamically intercept high-dimensional 4D activation tensors (B×C×H×W) and deflect adversarial noise into an orthogonal complement nullspace manifold. Constructed via singular value decomposition (SVD) across uncorrupted baseline calibration maps, the rank-K projection operator P_|| = U_K U_K^T attenuates non-subspace noise components without modifying underlying neural parameters. Executed under strict IEEE 754 Float64 double-precision numerical boundaries, the interception hook bounds the residual noise floor to ≤10^-12. Empirical evaluation over 10^6 continuous verification cycles demonstrates a runtime SLA overhead of +0.42 ms per batch, establishing an efficient, mathematically bounded defense layer for real-time inference pipelines.
Epistemological position. This work adopts a critical-rationalist stance (Popper, 1959): every theoretical claim is stated as a conjecture subject to empirical falsification, with quantitative rejection thresholds fixed a priori. Every mathematical guarantee is derived from expli…
This repository contains the datasets and source code associated with Georgiev et al., Science Advances (2026). https://doi.org/10.1126/sciadv.aec5080 To get started with the software, visit our GitHub repository. The software provides both a graphical user interface (GUI) and a…
Surveillance anomaly detection systems built around a single monolithic deep network are difficult to interpret, brittle to distribution shift, and offer operators no rationale on which to act. This paper presents VisionGuard, an explainable deep learning framework that reorganiz…
This preprint presents an empirical software engineering study on resolving main-thread performance bottlenecks in browser-based spatial computing. Abstract Real-time Retrieval-Augmented Generation (RAG) pipelines increasingly stream high-dimensional vector embeddings into browse…
Abstract- The intersection of Artificial Intelligence (AI), learning analytics, and digital twin is revolutionising higher education into an intelligent, data-driven, and ultimately personalised educational ecosystem. In contrast with today’s Learning Management Systems (LMSs), w…
Julia implementation of a scaled projection neural network for quasi-variational inequalities with state-dependent constraint set S(x) = m(x) + S and fixed symmetric positive-definite matrix M. Integrates the continuous-time dynamics dx/dt = lambda * [P_{S(x),M^{-1}}(x - alpha *…