Maya-CL: Nociceptive Metaplasticity and Vairagya-Governed Heterosynaptic Decay for Continual Learning in Spiking Neural Networks
Venkatesh Swaminathan · Zenodo (CERN European Organization for Nuclear Research) · 2026
Version 2 (2026-08-04). Revised to the NLL Universal Paper Format v6. The original text is retained in full; nothing has been deleted. Corrections appear as marked blocks placed at the section that carries the claim, and each one states what the paper said, what the data show, the corrected claim, why it happened, and what still stands. This version corrects a broken metric and an inverted abstract claim.FWT was computed against a chance level hardcoded at 10% in a 2-class protocol where chance is 50%, so the reported +40.00% and the +2.11 pp improvement derived from it are artefacts and true FWT is 0.00%; and the full method underperforms plain SGD on both metrics (AA 62.38 vs 64.57, BWT −30.55 vs −28.25), which the abstract had stated as a gain.The AA and BWT measurements, the ablation structure, and the honest report that Lability-only underperforms baseline all stand. Everything below this line is the original description from version 1. It is retained unchanged for the record. Where it conflicts with the corrections above, the corrections stand. Two framings it repeats have since been withdrawn in full — the Bhaya Quiescence Law and the Buddhi S-Curve. Both are addressed in Maya-Meta P1, now superseded, and in the self-audit of Maya-Meta P2. Maya-CL is a spiking neural network continual learning system designed to remember previously learned tasks without forgetting them — a longstanding challenge in AI — by drawing on biological principles of pain-driven plasticity and detachment-governed synaptic protection derived from Advaita Vedantic philosophy. Built on a fixed-weight convolutional architecture trained sequentially on five visual classification tasks without any replay memory or task identity cues, Maya-CL translates the Vedantic concept of Vairagya (non-attachment) into a gradient masking mechanism that selectively shields important synaptic weights from being overwritten by new learning. Across the Split-CIFAR-10 Task-Incremental Learning benchmark, full Maya-CL achieves an Average Accuracy of 62.38% and Forward Transfer of +40.00%, narrowing the backward transfer gap to baseline SGD by 2.30 percentage points and improving forward transfer by 2.11 percentage points — with ablation experiments providing the first quantitative isolation of heterosynaptic protection in a standard SNN continual learning setting. This work is part of the Maya Research Series from Nexus Learning Labs, Bengaluru, India — a 13-paper programme implementing the Advaita Vedantic Antahkarana as computational primitives in spiking neural networks — with the series-wide empirical constants of Bhaya Quiescence Law (β* ≤ 0.32%) and Buddhi S-Curve Determinism (R² = 1.0000) confirmed in this paper. Series: Part of the Maya Research Series — 13 papers implementing the Advaita Vedantic Antahkarana as computational primitives in spiking neural networks. Bhaya Quiescence Law (β* ≤ 0.32%) and Buddhi S-Curve Determinism (R²=1.0000) confirmed across all papers. Links: GitHub Repository (private — to request access: email [email protected] with subject Code Access Request — Maya-CL and your research context) | Full Series Index — venky2099.github.io Nexus Learning Labs, Bengaluru · UDYAM-KR-02-0122422 · BHASKAR IN-0526-9452JSORCID: 0000-0002-3315-7907 · VAIRAGYA_DECAY_RATE = 0.002315 — an ORCID-derived provenance mark, not an experimental parameter. Each paper's disclosure block states whether it reached a result in that paper.Canary: MayaNexusVS2026NLL_Bengaluru