METACOGNITIVE REGIME-SWITCHED SPIKING NEURAL NETWORK WITH CONTRASTIVE MEMORY CONSOLIDATION
DOI:
https://doi.org/10.62985/j.huit_ojs.vol26.no2E.423Keywords:
Spiking neural networks, metacognition, regime-switching, continual learning, memory consolidation, N-MNISTAbstract
Continual learning in Spiking Neural Networks (SNNs) remains a significant challenge due to the plasticity-stability dilemma, often leading to catastrophic forgetting. In this paper, we propose a novel computational architecture titled Metacognitive Regime-Switched SNN (MRS-SNN), which integrates a metacognitive control mechanism into a Complementary Learning Systems framework. The core of our architecture is the Regime-Switching Markov (RMS) controller, acting as a metacognitive agent [8] that monitors prediction errors in real-time to modulate gating signals. This allows the system to autonomously switch between a "Plasticity" mode (upon novelty detection) and a "Stability" mode (for familiar data). To address the cold-start problem and weight saturation inherent in noisy neuromorphic data, we introduce a Contrastive Memory Consolidation learning rule. This rule combines Gated Trace Learning with Subtractive Normalization, enabling sharp, one-shot feature extraction. Experimental validation on the N-MNIST dataset confirms that MRS-SNN maintains superior reliability across sequential tasks, with accuracy metrics reaching 83% for old tasks and 99% for new tasks (Global Avg: ~93%). Furthermore, through a simulated Wake-Sleep cycle, the system successfully consolidates memories from a short-term store (artificial Hippocampus) to a long-term store (artificial Neocortex), offering a promising solution for self-adaptive Edge AI systems.
References
[1] P. A. Merolla et al., “A million spiking-neuron integrated circuit with a scalable communication network and interface,” Science, vol. 345, no. 6197, pp. 668–673, 2014, doi: https://doi.org/10.1126/science.1254642.
[2] M. Davies et al., “Loihi: A neuromorphic manycore processor with on-chip learning,” IEEE Micro, vol. 38, no. 1, pp. 82–99, 2018, doi: https://doi.org/10.1109/mm.2018.112130359.
[3] J. Kirkpatrick et al., “Overcoming catastrophic forgetting in neural networks,” Proceedings of the National Academy of Sciences of the United States of America, vol. 114, no. 13, pp. 3521–3526, 2017, doi: https://doi.org/10.1073/pnas.1611835114.
[4] F. Zenke and S. Ganguli, “Supervised learning in multilayer spiking neural networks,” Neural Computation, vol. 30, no. 6, pp. 1514–1541, 2018, doi: https://doi.org/10.1162/neco_a_01086.
[5] E. O. Neftci, H. Mostafa, and F. Zenke, “Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based optimization to spiking neural networks,” IEEE Signal Processing Magazine, vol. 36, no. 6, pp. 51–63, 2019, doi: https://doi.org/10.1109/MSP.2019.2931595.
[6] J. L. McClelland, B. L. McNaughton, and R. C. O’Reilly, “Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory,” Psychological Review, vol. 102, no. 3, pp. 419–457, 1995, doi: https://doi.org/10.1037/0033-295X.102.3.419.
[7] D. Kumaran, D. Hassabis, and J. L. McClelland, “What learning systems do intelligent agents need? Complementary learning systems theory updated,” Trends in Cognitive Sciences, vol. 20, no. 7, pp. 512–534, 2016, doi: https://doi.org/10.1016/j.tics.2016.05.004
[8] S. M. Fleming and H. Lau, “How to measure metacognition,” Frontiers in Human Neuroscience, vol. 8, Art. no. 443, 2014, doi: https://doi.org/10.3389/fnhum.2014.00443.
[9] G. Orchard, A. Jayawant, G. K. Cohen, and N. Thakor, “Converting static image datasets to spiking neuromorphic datasets using saccades,” Frontiers in Neuroscience, vol. 9, Art. no. 437, 2015, doi: https://doi.org/10.3389/fnins.2015.00437.
