Maya-Manas: Oscillatory Thalamo-Cortical Gating for Class-Incremental Learning in Affective Spiking Neural Networks

We introduce Maya-Manas, the seventh paper in the Maya Research Series grounding spiking neural network (SNN) continual learning in Advaita Vedantic cognitive philosophy. Manas (मनस्) — the oscillating, doubting, sensory-receiving mind — is implemented as an Oscillatory Leaky Integrate-and-Fire (O-LIF) gate on the fc1 layer, tracing a half-cycle cosine threshold descent from Vikalpa (maximum suppression, t=0) to Sankalpa (full receptivity, t=3) within each forward pass. Only signals salient enough to spike at peak threshold earn amplified Vairagya protection via the Manas-GANE intersection. A 5-condition ablation on Split-CIFAR-100 Class-Incremental Learning (10 tasks, seed=42) confirms: Full Maya-Manas achieves AA=15.19% and BWT=−50.91%, a gain of +0.84pp AA and +1.77pp BWT over the P6 Maya-Chitta baseline — the best backward transfer in the Maya series to date. The Bhaya Quiescence Law is confirmed for the seventh consecutive paper. The negative control (B=A) confirms that structure without oscillation contributes zero gain.