本論文提出一套名為「SemaSNN-CL」的仿生持續學習框架，旨在解決深度神經網路中面臨的災難性遺忘問題。本研究模擬生物大腦互補學習系統之協同機制，在不依賴舊資料回放的前提下，實現穩健的終身學習。
首先，本研究核心在於模擬海馬迴中由輸入驅動的閘控機制，引入「動態路由」調控策略。本研究提出影像幾何共振機制與語意交互共振機制，依據輸入特徵即時調節訊號傳遞路徑，將高階特徵轉化為具備空間選擇性的閘控訊號，藉此選擇性活化特定神經元以實現精確的模式分離。其次，將神經科學的突觸標記與捕捉(Synaptic Tagging and Capture, STC)假說轉化為演算法。提出「新穎性驅動」更新策略，依據神經元對新資訊的預期誤差定義可量化的新穎性指標，藉此進行分子標記與結構性固化，主動保護關鍵的記憶痕跡，同時保留閒置突觸的可塑性以適應新任務。最後，經實驗證實此仿生機制能有效在記憶穩定性與學習可塑性之間取得平衡，大幅降低跨任務的遺忘率，並藉由脈衝神經網路稀疏運算顯著降低推論能耗。

This thesis proposes SemaSNN-CL, a brain-inspired continual learning framework designed to mitigate catastrophic forgetting in deep neural networks. By emulating the cooperative principles of the Complementary Learning Systems (CLS) in the brain, SemaSNN-CL enables robust lifelong learning without replaying past data.
At its core, SemaSNN-CL instantiates hippocampal-style input-driven gating as a dynamic routing mechanism. We introduce geometric resonance (for vision) and semantic interactive resonance (for sequence modeling) to convert high-level representations into spatially selective gating signals that regulate information flow in real time, selectively activating task-relevant neurons and promoting precise pattern separation. In addition, we translate the neuroscience hypothesis of Synaptic Tagging and Capture (STC) into an algorithmic consolidation process. Specifically, we propose a novelty-driven update rule that quantifies novelty via prediction-error–based signals, triggering tagging and structural consolidation to protect critical memory traces while maintaining the plasticity of idle synapses for future adaptation.
Extensive experiments demonstrate that SemaSNN-CL achieves a favorable trade-off between memory stability and learning plasticity, substantially reducing forgetting across tasks under replay-free settings. Moreover, leveraging the sparse computation of spiking neural networks (SNNs), the proposed framework significantly reduces inference energy consumption while preserving competitive performance.

Table of Contents
Acknowledgements	i
Abstract	xi
Table of Contents	xiii
List of Figures	xvi
List of Tables	xvii
Chapter 1 Introduction	1
1.1 Motivation	1
1.2 Objectives	3
1.3 Research Scope	4
1.4 Contributions	5
1.5 Thesis Organization	7
Chapter 2. Related Work	8
2.1 Catastrophic Forgetting in Continual Learning and Related Approaches	8
2.1.1 Regularization-based Methods	9
2.1.2 Replay and Distillation Strategies	9
2.1.3 Parameter Isolation and Bio-inspired Dynamic Architectures	10
2.1.4 Parameter-Efficient Fine-Tuning and Prompting Strategies	11
2.2 Biological Mechanisms and Biomimetic Learning Theories	12
2.2.1 Hippocampal Synaptic Pathways and Pattern Separation	12
2.2.2 Memory Allocation and Inhibitory Competition	12
2.2.3 Novelty Detection and Synaptic Tagging and Capture	13
2.3 Dynamic Routing and Feature Gating	14
2.4 Spiking Neural Networks and Temporal Filtering Mechanisms	15
2.5 Semantic Representations and Modular Adaptation in Large Language Models	16
2.6 Summary of Related Work	17
Chapter 3. Framework Overview of SemaSNN-CL	18
3.1 Theoretical Core of the SemaSNN-CL Architecture	18
3.2 Memory Engrams and Spatial Partitioning: Pattern Separation	20
3.3 Universal Neural Dynamics: Physical Filtering at the Cell Membrane	21
3.4 Plasticity Modulation: Weight Consolidation and Gradient Blocking	23
3.5 Conclusions	24
Chapter 4. SemaSNN-CL for Continual Image Classification	26
4.1 Hippocampal System Modeling: Pattern Separation and Structural Plasticity	27
4.1.1 Geometric Resonance and Conditional Pattern Separation	28
4.1.2 Dynamic Evolution and Consolidation of Memory Engrams	30
4.1.3 Physical Addressing and De-inhibition	32
4.2 Neocortical Dynamics: Temporal Filtering with PLIF Neurons	34
4.3 STC Control Loop: From Dopamine Signals to Molecular Consolidation	36
4.3.1 Neuronal Pathway: Novelty Detection	37
4.3.2 Synaptic Pathway: Importance Quantification	37
4.3.3 Gating Logic: Saliency-Based Gating	38
4.3.4 State and Actuation: Gradient Masking and Consolidation	39
4.4 Integrated Workflow: A Small Neocortex and a Large Hippocampus	40
4.5 Summary	42
Chapter 5. SemaSNN-CL for Knowledge Learning in LLMs	43
5.1 Dentate Gyrus Simulation: CBG-Based Pattern Separation	44
5.1.1 Limitations of Linear Projection: Feature Superposition	44
5.1.2 Compact Bilinear Interaction: Second-Order Enhancement	45
5.1.3 Rank Projection: Sketch to Low-Rank Driv	47
5.2 Soft-LIF Temporal Inertial Filtering	47
5.2.1 Leaky Integration: Physical Inertia of Semantic Flow	48
5.2.2 Graded Potential: High-Resolution Semantic Encoding	49
5.2.3 Soft Reset: Non-Saturating Sparsity for Parameter Isolation	50
5.3 LoRA Implementation with Neuromodulatory Control	51
5.3.1 Dynamic Rank-Level Gating	52
5.3.2 Orthogonal Subspace Initialization	53
5.3.3 Regularization Objectives	54
Chapter 6. Experiments and Ablation Studies	56
6.1 Experimental Setup and Datasets	56
6.1.1 Image Classification Datasets	56
6.1.2 Language Model Datasets	58
6.1.3 Evaluation Metrics	60
6.2 Vision Experiments: Stability of Microscopic Geometry	62
6.2.1. Vision Experiment I: MNIST	63
6.2.2 Vision Experiment II: CIFAR-100	64
6.2.3 Vision Experiment III: Tiny-ImageNet	66
6.3 Language Experiments: Isolation of Macroscopic Semantics	70
6.3.1 Language Experiment I: MMLU	70
6.3.2 Language Experiment II: Synthetic Keyword QA	73
6.4 Ablation Studies and Mechanism Validation	78
6.4.1 Ablation Study on Visual Models	78
6.4.2 Ablation Study on Language Models	80
6.5 Visualization and Interpretability Analysis	83
6.5.1 Neuronal Gating Signals in the Vision Model	84
6.5.2 Neuronal Gating Signals in the Language Models	88
6.6 Parameter Efficiency and Computational Cost Analysis	93
6.6.1 Definition of Computational Efficiency Metrics	93
6.6.2 Performance Cost Analysis of the SemaSNN-CL Visual Model	94
6.6.3 Performance Cost Analysis of the SemaSNN-CL Language Model	96
6.7 Summary	97
Chapter 7. Conclusion and Future Work	100
7.1 Conclusions	100
7.2 Future Work	102
References	104

 
List of Figures
Fig 1. 1 Schematic of hippocampal memory formation and pattern separation	2
Fig 3. 1 Overall Architecture of SemaSNN-CL	18
Fig 4. 1 SemaSNN-CL for Image Classification Architecture	27
Fig 4. 2 SemaSNN-CL for Resonance and Pattern Separation in Visual Perception	29
Fig 4. 3 Normalized Geometric Drift Strategy	31
Fig 4. 4 Physical Addressing in the Hippocampal Module	32
Fig 4. 5 Detailed computational flow of a PLIF neuron	34
Fig 4. 6 Detailed workflow of the STC control loop	37
Fig 4. 7 Overall visual architecture of SemaSNN-CL	40
Fig 5. 1 Overall architecture of SemaSNN-CL for language models	44
Fig 5. 2 Trainable CBG Gate	45
Fig 5. 3 Soft-LIF Node Architecture	48
Fig 5. 4 Detailed architecture of SemaSNN-CL for language models	52
Fig 6. 1 Partitioning of the MNIST handwritten digit dataset	57
Fig 6. 2 Partitioning of the CIFAR-100 dataset	57
Fig 6. 3 Partitioning of the Tiny-ImageNet dataset	58
Fig 6. 4 Format of the MMLU dataset	58
Fig 6. 5 Format of the synthetic keyword question-answering dataset	60
Fig 6. 6 Average accuracy on MNIST (10 tasks)	63
Fig 6. 7 Average accuracy on Cifar100 (20 tasks)	65
Fig 6. 8 Average accuracy on Tiny-ImageNet (40 tasks)	67
Fig 6. 9 Task-wise accuracy heatmap across all stages on Tiny-ImageNet (40 tasks)	69
Fig 6. 10 Average accuracy on MMLU (100×10 tasks)	72
Fig 6. 11 Average accuracy on MMLU (10×57 tasks)	72
Fig 6. 12 Average accuracy on Synthetic Keyword QA (10×10 tasks)	74
Fig 6. 13 Task-wise accuracy heatmap across all stages on SKQA 10×10-tasks	75
Fig 6. 14 Average accuracy on Synthetic Keyword QA (1×100 tasks)	76
Fig 6. 15 Task-wise accuracy heatmap across all stages on SKQA 1×100-tasks	77
Fig 6. 16 Average accuracy curves of the SemaSNN-CL visual ablation experiments	78
Fig 6. 17 Average accuracy curves of the SemaSNN-CL language ablation experiments	80
Fig 6. 18 Average gating signals of the first 15 Tiny-ImageNet classes	84
Fig 6. 19 Gating signals of 5 images per class in Tiny-ImageNet Task 1	85
Fig 6. 20 t-SNE visualization of gating signals for 40 Tiny-ImageNet classes	87
Fig 6. 21 Average low-rank gating heatmap across five MMLU domains	89
Fig 6. 22 Low-rank gating heatmap across five MMLU domains (100 samples each)	90
Fig 6. 23 t-SNE visualization of gating signals across five MMLU domains	91
Fig 6. 24 Computational efficiency of SemaSNN-CL versus ANN on Tiny-ImageNet	95
Fig 6. 25 Performance and cost comparison between SemaSNN-CL and Standard LoRA	96
List of Tables
Table 6. 1 Overall performance summary of the MNIST 10-tasks vision experiment	63
Table 6. 2 Overall performance summary of the CIFAR-100 20-tasks vision experiment	65
Table 6. 3 Overall performance summary of the Tiny-ImageNet 40-tasks vision experiment	67
Table 6. 4 Overall performance summary of the MMLU 100×10-tasks language experiment	72
Table 6. 5 Overall performance summary of the MMLU 10×57-tasks language experiment	73
Table 6. 6 Overall performance summary of the SKQA 10×10-tasks language experiment	74
Table 6. 7 Overall performance summary of the SKQA 1×100-tasks language experiment	76
Table 6. 8 Summary of metrics for the SemaSNN-CL visual ablation study	79
Table 6. 9 Summary of metrics for the SemaSNN-CL language ablation study	81

References
[1]	Zhao WX, et al. A survey of large language models. arXiv preprint arXiv:230318223, (2023).
[2]	Qu H, Rahmani H, Xu L, Williams B, Liu J. Recent advances of continual learning in computer vision: An overview. IET Computer Vision 19, e70013 (2025).
[3]	Y. Luo, Z. Yang, F. Meng, Y. Li, J. Zhou, and Y. Zhang, "An empirical study of catastrophic forgetting in large language models during continual fine-tuning," arXiv preprint arXiv:2308.08747, 2023.
[4]	M. Mermillod, A. Bugaiska, and P. Bonin, "The stability-plasticity dilemma: Investigating the continuum from catastrophic forgetting to age-limited learning effects," Front. Psychol., vol. 4, p. 504, 2013.
[5]	E. J. Hu et al., "LoRA: Low-rank adaptation of large language models," in Proc. Int. Conf. Learn. Represent. (ICLR), 2022.
[6]	G. I. Parisi, R. Kemker, J. L. Part, C. Kanan, and S. Wermter, "Continual lifelong learning with neural networks: A review," Neural Netw., vol. 113, pp. 54–71, 2019.
[7]	Treves and E. T. Rolls, "Computational analysis of the role of the hippocampus in memory," Hippocampus, vol. 4, no. 3, pp. 374–391, Jun. 1994.
[8]	J. K. Leutgeb, S. Leutgeb, M.-B. Moser, and E. I. Moser, "Pattern separation in the dentate gyrus and CA3 of the hippocampus," Science, vol. 315, no. 5814, pp. 961–966, Feb. 2007.
[9]	T. J. McHugh et al., "Dentate gyrus NMDA receptors mediate rapid pattern separation in the hippocampal network," Science, vol. 317, no. 5834, pp. 94–99, Jul. 2007.
[10]	W. Maass, "Networks of spiking neurons: The third generation of neural network models," Neural Networks, vol. 10, no. 9, pp. 1659–1671, 1997.
[11]	M. Rabinovich, R. Huerta, and G. Laurent, "Transient dynamics for neural processing," Science, vol. 321, no. 5885, pp. 48–50, 2008.
[12]	(mnist)Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proc. IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.
[13]	(cifar100)A. Krizhevsky, "Learning multiple layers of features from tiny images," Univ. of Toronto, Toronto, ON, Canada, Tech. Rep., 2009.
[14]	(tinyimagenet)J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, "ImageNet: A large-scale hierarchical image database," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2009, pp. 248–255.
[15]	(mmlu)D. Hendrycks et al., "Measuring massive multitask language understanding," in Proc. Int. Conf. Learn. Represent. (ICLR), 2021.
[16]	(PLIF)W. Fang, Z. Yu, Y. Yan, D. Masquelier, T. Huang, and Y. Tian, "Incorporating learnable membrane time constant to enhance learning of spiking neural networks," in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 2661–2671.
[17]	(STC)U. Frey and R. G. M. Morris, "Synaptic tagging and long-term potentiation," Nature, vol. 385, no. 6616, pp. 533–536, 1997.
[18]	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," Psychol. Rev., vol. 102, no. 3, pp. 419–457, Jul. 1995.
[19]	J. Kirkpatrick et al., "Overcoming catastrophic forgetting in neural networks," Proc. Nat. Acad. Sci. U.S.A., vol. 114, no. 13, pp. 3521–3526, Mar. 2017.
[20]	F. Zenke, B. Poole, and S. Ganguli, "Continual learning through synaptic intelligence," in Proc. 34th Int. Conf. Mach. Learn. (ICML), 2017, pp. 3987–3995.
[21]	S. Fusi and L. F. Abbott, "Limits on the memory storage capacity of bounded synapses," Nature Neurosci., vol. 10, no. 4, pp. 485–493, Apr. 2007.
[22]	D. Lopez-Paz and M. A. Ranzato, "Gradient episodic memory for continual learning," in Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 30, 2017.
[23]	G. M. van de Ven, H. T. Siegelmann, and A. S. Tolias, "Brain-inspired replay for continual learning with artificial neural networks," Nature Commun., vol. 11, no. 1, Art. no. 4069, Aug. 2020.
[24]	S.-A. Rebuffi, A. Kolesnikov, G. Sperl, and C. H. Lampert, "iCaRL: Incremental classifier and representation learning," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 2001–2010.
[25]	A. Mallya and S. Lazebnik, "PackNet: Adding multiple tasks to a single network by iterative pruning," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2018, pp. 7765–7773.
[26]	A. Mallya, D. Davis, and S. Lazebnik, "Piggyback: Adapting a single network to multiple tasks by learning to mask weights," in Proc. Eur. Conf. Comput. Vis. (ECCV), 2018, pp. 67–82.
[27]	F.-Y. Wang, D.-W. Zhou, H.-L. Ye, and Z.-H. Zhan, "FOSTER: Feature boosting and compression for class-incremental learning," in Proc. Eur. Conf. Comput. Vis. (ECCV), 2022, pp. 398–414.
[28]	N. Y. Masse, G. D. Grant, and D. J. Freedman, "Alleviating catastrophic forgetting using context-dependent gating and synaptic stabilization," Proc. Nat. Acad. Sci. U.S.A., vol. 115, no. 44, pp. E10467–E10475, Oct. 2018.
[29]	Q. Shi et al., "Hybrid neural networks for continual learning inspired by corticohippocampal circuits," Nature Commun., vol. 16, no. 1, Art. no. 745, 2025.
[30]	X. Wang et al., "Orthogonal subspace learning for language model continual learning," in Findings of the Assoc. Comput. Linguist.: EMNLP 2023, Dec. 2023, pp. 245–258.
[31]	S. Dou et al., "LoRAMoE: Alleviating world knowledge forgetting in large language models via moe-style plugin," arXiv preprint arXiv:2312.09979, 2023.
[32]	T. Brown et al., "Language models are few-shot learners," in Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 1877–1901.
[33]	Z. Wang et al., "Learning to prompt for continual learning," in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 139–149.
[34]	P. Lewis et al., "Retrieval-augmented generation for knowledge-intensive NLP tasks," in Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 9459–9474.
[35]	D. Marr, "Simple memory: a theory for archicortex," Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, vol. 262, no. 841, pp. 23–81, 1971.
[36]	J. C. R. Whittington et al., "The Tolman-Eichenbaum Machine: Unifying space and relational memory for generalisation," Cell, vol. 183, no. 5, pp. 1249–1263, 2020.
[37]	X. Liu, S. Ramirez, P. T. Pang, C. B. Puryear, A. Govindarajan, K. M. Deisseroth, and S. Tonegawa, "Optogenetic stimulation of a hippocampal engram activates fear memory recall," Nature, vol. 484, no. 7394, pp. 381–385, 2012.
[38]	S. A. Josselyn and S. Tonegawa, "Memory engrams: Recalling the past and imagining the future," Science, vol. 367, no. 6473, Art. no. eaaw4325, 2020.
[39]	T. P. Vogels, H. Sprekeler, F. Zenke, C. Clopath, and W. Gerstner, "Inhibitory plasticity balances excitation and inhibition in sensory pathways and memory networks," Science, vol. 334, no. 6062, pp. 1569–1573, 2011.
[40]	S. Sadeh and C. Clopath, "Excitatory–inhibitory balance in neuronal networks: from mechanisms to functions," Nature Reviews Neuroscience, vol. 22, no. 1, pp. 21–37, 2021.
[41]	J. E. Lisman and A. A. Grace, "The hippocampal-VTA loop: controlling the entry of information into long-term memory," Neuron, vol. 46, no. 5, pp. 703–713, 2005.
[42]	A. J. Duszkiewicz, C. G. McNamara, T. Takeuchi, and L. Genzel, "Novelty and dopaminergic modulation of memory persistence: a tale of two systems," Trends in Neurosciences, vol. 47, no. 2, pp. 102–114, 2024.
[43]	Y. Li et al., "Dendritic computing in spiking neural networks for synaptic tagging and capture," Nature Communications, vol. 14, no. 1, Art. no. 7945, 2023.
[44]	B. A. Olshausen and D. J. Field, "Emergence of simple-cell receptive field properties by learning a sparse code for natural images," Nature, vol. 381, no. 6583, pp. 607–609, 1996.
[45]	S. Grossberg, "Adaptive Resonance Theory: How a brain learns to consciously attend, learn, and recognize a changing world," Neural Networks, vol. 37, pp. 1–47, 2013.
[46]	J. Frankle and M. Carbin, "The lottery ticket hypothesis: Finding sparse, trainable neural networks," in International Conference on Learning Representations (ICLR), 2019.
[47]	S. Sabour, N. Frosst, and G. E. Hinton, "Dynamic routing between capsules," in Advances in Neural Information Processing Systems (NeurIPS), vol. 30, pp. 3856–3866, 2017.
[48]	N. Shazeer et al., "Outrageously large neural networks: The sparsely-gated mixture-of-experts layer," in International Conference on Learning Representations (ICLR), 2017.
[49]	C. Riquelme et al., "Scaling vision with sparse mixture of experts," in Advances in Neural Information Processing Systems (NeurIPS), vol. 34, pp. 8583–8595, 2021.
[50]	Y. Gao, O. Beijbom, N. Zhang, and T. Darrell, "Compact bilinear pooling," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 317–326.
[51]	W. Gerstner, W. M. Kistler, R. Naud, and L. Paninski, Neuronal Dynamics: From Single Neurons to Networks and Models of Cognition. Cambridge, U.K.: Cambridge University Press, 2014.
[52]	W. Fang, Z. Yu, Y. Chen, T. Masquelier, T. Huang, and Y. Tian, "Deep residual learning in spiking neural networks," in Advances in Neural Information Processing Systems (NeurIPS), vol. 34, pp. 21056–21069, 2021.
[53]	M. Horowitz, "1.1 Computing's energy problem (and what we can do about it)," in 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), 2014, pp. 10–14.
[54]	Z. Zhou, Y. Zhu, C. He, Y. Wang, S. Yan, Y. Tian, and L. Yuan, "Spikformer: A spiking transformer," in International Conference on Learning Representations (ICLR), 2024.
[55]	R.-J. Zhu, Q. Zhao, G. Li, and J. K. Eshraghian, "SpikeGPT: Generative pre-trained language model with spiking neural networks," arXiv preprint arXiv:2302.13939, 2023.
[56]	A. Vaswani et al., "Attention is all you need," in Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017.
[57]	R. Bommasani et al., "On the opportunities and risks of foundation models," arXiv preprint arXiv:2108.07258, 2021.
[58]	A. Yang et al., "Qwen2.5 technical report," arXiv preprint arXiv:2412.15115, 2024.
[59]	A. Dubey et al., "The Llama 3 herd of models," arXiv preprint arXiv:2407.21783, 2024.
[60]	A. Aghajanyan, L. Zettlemoyer, and S. Gupta, "Intrinsic dimensionality explains the effectiveness of language model fine-tuning," in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL), 2021, pp. 7319–7328.
[61]	J. Pfeiffer et al., "Modular deep learning," arXiv preprint arXiv:2302.11529, 2023.
[62]	D. Kumaran, D. Hassabis, and J. L. McClelland, "What learning systems do intelligent agents need? Complementary learning systems theory updated," Trends Cogn. Sci., vol. 20, no. 7, pp. 512–534, 2016.
[63]	H. K. Hartline and F. Ratliff, "Inhibitory interaction of receptor units in the eye of Limulus," J. Gen. Physiol., vol. 40, no. 3, pp. 357–376, Jan. 1957.
[64]	A. L. Hodgkin and A. F. Huxley, "A quantitative description of membrane current and its application to conduction and excitation in nerve," J. Physiol., vol. 117, no. 4, pp. 500–544, 1952.
[65]	W. Rall, "Distinguishing theoretical synaptic potentials computed for different soma-dendritic distributions of synaptic input," J. Neurophysiol., vol. 30, no. 5, pp. 1138–1168, 1967.
[66]	R. C. O'Reilly and K. A. Norman, "Hippocampal and neocortical contributions to memory: Advances in the complementary learning systems framework," Trends Cogn. Sci., vol. 6, no. 12, pp. 505–510, 2002.
[67]	J. J. Knierim and J. P. Neunuebel, "Tracking the flow of hippocampal computation: Pattern separation, pattern completion, and attractor dynamics," Neurobiol. Learn. Mem., vol. 129, pp. 38–49, 2016.
[68]	D. H. Hubel and T. N. Wiesel, "Receptive fields, binocular interaction and functional architecture in the cat's visual cortex," J. Physiol., vol. 160, no. 1, pp. 106–154, 1962.
[69]	E. M. Izhikevich, "Which model to use for cortical spiking neurons?" IEEE Trans. Neural Netw., vol. 15, no. 5, pp. 1063–1070, 2004.
[70]	S. Grossberg, "Contour enhancement, short-term memory, and constancies in reverberating neural networks," Stud. Appl. Math., vol. 52, no. 3, pp. 213–257, 1973.
[71]	V. F. Castellucci and E. R. Kandel, "A quantal analysis of the synaptic depression underlying habituation of the gill-withdrawal reflex in Aplysia," Proc. Nat. Acad. Sci. U.S.A., vol. 71, no. 12, pp. 5004–5008, 1974.
[72]	S. Marsland, J. Shapiro, and U. Nehmzow, "A self-organising network that grows when required," Neural Netw., vol. 15, no. 8–9, pp. 1041–1058, 2002.
[73]	W. Fang, Z. Yu, Y. Yan, D. Masquelier, T. Huang, and Y. Tian, "Incorporating learnable membrane time constant to enhance learning of spiking neural networks," in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 2661–2671.
[74]	M. Minsky and S. Papert, Perceptrons: An Introduction to Computational Geometry. Cambridge, MA, USA: MIT Press, 1969.
[75]	M. E. Hasselmo, "Neuromodulation: Acetylcholine and memory consolidation," Trends Cogn. Sci., vol. 3, no. 9, pp. 351–359, 1999.
[76]	H. Bailey, M. Giustetto, Y. Y. Huang, D. O. Hawkins, and E. R. Kandel, "Is heterosynaptic modulation essential for stabilizing massive synaptic strengthening?" Nat. Rev. Neurosci., vol. 1, no. 1, pp. 11–20, 2000.
[77]	Liao, N. A. Hessler, and R. Malinow, "Activation of postsynaptically silent synapses during pairing-induced LTP in CA1 region of hippocampal slice," Nature, vol. 375, no. 6530, pp. 400–404, 1995.
[78]	G. M. van de Ven and A. S. Tolias, "Three scenarios for continual learning," arXiv preprint arXiv:1904.07734, 2019.
[79]	Radford et al., "Learning transferable visual models from natural language supervision," in Proc. Int. Conf. Mach. Learn. (ICML), 2021, pp. 8748–8763.
[80]	R. Schwartz, J. Dodge, N. A. Smith, and O. Etzioni, "Green AI," Commun. ACM, vol. 63, no. 12, pp. 54–63, 2020.