1.Medical Innovation Research Department of PLA General Hospital, Key Laboratory of Biomedical Engineering and Translational Medicine, Ministry of Industry and Information Technology, Beijing 100853, China
2.School of Biological and Medical Engineering, Beihang University, Beijing 100191, China
Based on a deep understanding of the visual encoding neurophysiological mechanisms between the retina, lateral geniculate body, and visual cortex of the brain, we have constructed a novel neural network-based brain-inspired intelligent unit architecture, laying a solid foundation for the implementation of large-scale integrated neural network-based brain-inspired models. This information theory foundation for our understanding of the expression and computation of brain sensory encoding is formed by the architecture of brain-inspired intelligent units. This article delves into the training methods, strategies, and specific algorithm examples of brain-inspired models, and proposes a comprehensive strategy that combines the redundancy reduction principle of sensory data flow expression and computation, self-organizing feature mapping, and recurrent oscillation synchronization mechanism, aiming to improve the biological rationality and interpretability of brain-inspired models, as well as efficiently and quickly mimic complex brain functions.
BarlowHB. Sensory mechanisms,the reduction of redundancy,and intelligence[C]//Proceedings of the symposium on the mechanization of thought processes. London:National Physical Laboratory,1959.
[2]
BarlowHB. Possible principles underlying the transformations of sensory messages[M/OL].
AtickJJ, LiZP, RedlichAN. Understanding retinal color coding from first principles[J]. Neural Comput,1992,4(4): 559-572.
[5]
AttneaveF. Some informational aspects of visual perception[J]. Psychol Rev,1954,61(3): 183-193.
[6]
WatanabeS. Information theoretical analysis of multivariate correlation[J]. IBM J Res Dev,1960,4(1): 66-82.
[7]
ZhaoHL, LiZN, SuWS,et al. Dynamic sparse coding-based value estimation network for deep reinforcement learning[J]. Neural Netw,2023,168: 180-193.
[8]
BowrenJ, Sanchez-GiraldoL, SchwartzO. Inference via sparse coding in a hierarchical vision model[J]. J Vis,2022,22(2): 19.
[9]
TangMY, WuYF. A blind source separation method based on bounded component analysis optimized by the improved beetle antennae search[J]. Sensors,2023,23(19): 8325.
[10]
AnsariS, AlatranyAS, AlnajjarKA,et al. A survey of artificial intelligence approaches in blind source separation[J]. Neurocomputing,2023,561:126895.
[11]
HyvärinenA, KhemakhemI, MoriokaH. Nonlinear independent component analysis for principled disentanglement in unsupervised deep learning[J]. Patterns (N Y),2023,4(10): 100844.
[12]
Mestre-BachG, GraneroR, Fernández-ArandaF,et al. Independent component analysis for Internet gaming disorder[J]. Dialogues Clin Neurosci,2023,25(1): 14-23.
HamamotoR, TakasawaK, MachinoH,et al. Application of non-negative matrix factorization in oncology:one approach for establishing precision medicine[J]. Brief Bioinform,2022,23(4): bbac246.
[15]
AonishiT, MaruyamaR, ItoT,et al. Imaging data analysis using non-negative matrix factorization[J]. Neurosci Res,2022,179: 51-56.
[16]
WuXF, ChengL, ZhangSF. Open set domain adaptation with entropy minimization[C]//Chinese Conference on Pattern Recognition and Computer Vision (PRCV). Cham:Springer,2020:29-41.
[17]
PoduvalP, OberoiG, VermaS,et al. BipNRL:mutual information maximization on bipartite graphs for node representation learning[C]//Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham:Springer,2023:728-743.
[18]
WanWQ, ChenJL, XieJS. MIM-Graph:a multi-sensor network approach for fault diagnosis of HSR Bogie bearings at the IoT edge via mutual information maximization[J]. ISA Trans,2023,139: 574-585.
[19]
WróbelS, TurekC, StępieńE,et al. Data integration through canonical correlation analysis and its application to OMICs research[J]. J Biomed Inform,2024,151: 104575.
[20]
NadalJP, PargaN. Nonlinear neurons in the low-noise limit:a factorial code maximizes information transfer[J]. Netw Comput Neural Syst,1994,5(4): 565-581.
[21]
MłynarskiWF, HermundstadAM. Efficient and adaptive sensory codes[J]. Nat Neurosci,2021,24(7): 998-1009.
[22]
KafashanM, JaffeAW, ChettihSN,et al. Scaling of sensory information in large neural populations shows signatures of information-limiting correlations[J]. Nat Commun,2021,12(1): 473.
[23]
AkhavanS, Soltanian-ZadehH. Blind separation of sparse sources from nonlinear mixtures[J]. Digit Signal Process,2021,118: 103220.
[24]
MallatS. Understanding deep convolutional networks[J]. Philos Trans A Math Phys Eng Sci,2016,374(2065): 20150203.
[25]
RumyantsevOI, LecoqJA, HernandezO,et al. Fundamental bounds on the fidelity of sensory cortical coding[J]. Nature,2020,580(7801): 100-105.
[26]
SteinfeldR, Tacão-MonteiroA, RenartA. Differential representation of sensory information and behavioral choice across layers of the mouse auditory cortex[J]. Curr Biol,2024,34(10):2200-2211.
[27]
LinskerR. Self-organization in a perceptual network[J]. Computer,1988,21(3): 105-117.
[28]
WuYN, SaiGL, DuanSY. Work-in-progress:accelerated matrix factorization by approximate computing for recommendation system[C]//2022 International Conference on Embedded Software (EMSOFT). IEEE,2022:1-2.
[29]
BromleyJ, GuyonI, LeCunY,et al. Signature verification using a “Siamese” time delay neural network[C]//Proceedings of the 6th International Conference on Neural Information Processing Systems. Denver:ACM,1993:737-744.
GrillJB, StrubF, AltchéF,et al. Bootstrap your own latent a new approach to self-supervised learning[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems. Vancouver:ACM,2020:21271-21284.
[33]
BardesA, PonceJ, LeCunY. VICReg:variance-invariance-covariance regularization for self-supervised learning[EB/OL].
[34]
CaronM, MisraI, MairalJ,et al. Unsupervised learning of visual features by contrasting cluster assignments[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems. Vancouver:ACM,2020:9912-9924.
[35]
JureZ, LiJ, IshanM,et al. Barlow twins:self-supervised learning via redundancy reduction[J/OL].
[36]
BarlowHB, TolhurstDJ. Why do you have edge detectors?[C]//Optical Society of America Annual Meeting. Albuquerque:Optica Publishing Group,1992.
[37]
HubelDH, WieselTN. Receptive fields and functional architecture of monkey striate cortex[J]. J Physiol,1968,195(1): 215-243.
[38]
MarrD, HildrethE. Theory of edge detection[J]. Proc R Soc Lond B Biol Sci,1980,207(1167): 187-217.
[39]
ZylberbergJ, MurphyJT, DeWeeseMR. A sparse coding model with synaptically local plasticity and spiking neurons can account for the diverse shapes of V1 simple cell receptive fields[J]. PLoS Comput Biol,2011,7(10): e1002250.
[40]
MijatovicG, AntonacciY, Loncar-TurukaloT,et al. An information-theoretic framework to measure the dynamic interaction between neural spike trains[J]. IEEE Trans Biomed Eng,2021,68(12): 3471-3481.
[41]
LarssonDT, MaityD, TsiotrasP. A generalized information-theoretic framework for the emergence of hierarchical abstractions in resource-limited systems[J]. Entropy,2022,24(6):809.
[42]
ChenKX, BeyelerM, KrichmarJL. Cortical motion perception emerges from dimensionality reduction with evolved spike-timing-dependent plasticity rules[J]. J Neurosci,2022,42(30): 5882-5898.
[43]
FradyEP, KleykoD, SommerFT. Variable binding for sparse distributed representations:theory and applications[J]. IEEE Trans Neural Netw Learn Syst,2023,34(5): 2191-2204.
[44]
LimaB, FlorentinoMM, FioraniM,et al. Cortical maps as a fundamental neural substrate for visual representation[J]. Prog Neurobiol,2023,224: 102424.
[45]
CavanaghSE, HuntLT, KennerleySW. A diversity of intrinsic timescales underlie neural computations[J]. Front Neural Circuits,2020,14: 615626.
[46]
ShepherdGMG, YamawakiN. Untangling the cortico-thalamo-cortical loop:cellular pieces of a knotty circuit puzzle[J]. Nat Rev Neurosci,2021,22(7): 389-406.
[47]
BalasB, SavilleA. Neural sensitivity to natural image statistics changes during middle childhood[J]. Dev Psychobiol,2021,63(5): 1061-1070.