关于稀疏编码在图像处理中的神经动力学分析

PDF(1168 KB)

振动与冲击 ›› 2018, Vol. 37 ›› Issue (22) : 17-21.

论文

卢颍霞，王如彬

作者信息 +

Nerve dynamics analysis on the sparse coding in image processing

LU Yingxia,WANG Rubin

Author information +

文章历史 +

摘要

初级视皮层V1区神经细胞采用稀疏编码的形式来有效表示自然场景，自然场景的稀疏编码模型可以解释V1区神经元的一些生理学性质，但是我们不知道这种编码模型是否可以通过生物学上的局部突触可塑性规则学习得到。由于生物神经网络中存在一种侧抑制现象即局部竞争，我们基于这种现象并利用突触局部可塑性规则建立了一个发放的神经网络动力学模型。对V1区细胞的感受野进行了仿真，同时利用模型得到的稀疏重构系数，对重构残差进行了讨论。研究表明利用稀疏编码可以得到V1区简单细胞的感受野，同时利用自然图像的输入说明了该动力学模型在生理学意义上的合理性。

Abstract

The sparse coding model for natural scene can explain some physiological properties of neurons in V1, but it is still unknown whether this coding model could be learned according to the biologically local synaptic plasticity rule.Due to the existence of the lateral inhibition phenomenon in the biological neural network, a neural network dynamic model was established based on this phenomenon and using the local synaptic plasticity.The simulation of the receptive field of V1 cells was carried out.At the same time, the reconstructed residuals were discussed by using the sparse reconstruction coefficients obtained by the model.The results show that the susceptibility field of V1 cells can be obtained by using the sparse coding, and the rationality of the kinetic model can be described by the input of natural images.

导出引用

卢颍霞，王如彬. 关于稀疏编码在图像处理中的神经动力学分析[J]. 振动与冲击, 2018, 37(22): 17-21

LU Yingxia,WANG Rubin. Nerve dynamics analysis on the sparse coding in image processing[J]. Journal of Vibration and Shock, 2018, 37(22): 17-21

参考文献

[1] Graham DJ, Field DJ. Sparse coding in the neocortex[J]. In: Kass JH, ed. Evolution of the nervous system, Vol III. Oxford: Academic Press,2007: 181–187.
[2] Lennie P. The cost of cortical computa-tion[J]. CurrBiol,2003,13:493–497.
[3] Hromadka T. Sparse representation of sounds in the unanesthetized auditory cortex[J]. PLoS Biol,2008, 6: 124–137.
[4] Vinje WE, Gallant JL. Natural stimulation of the nonclassical receptive field increases information transmission efficiency in v1[J]. J Neurosci ,2002,22:2904–2915.
[5] Vinje WE, Gallant JL. Sparse coding and decorrelation in primary visual cortex during natural vision[J]. Science, 2000,287: 1273–1276.
[6] Haider BA, Krause MR, Duque A, et al. Synaptic and network mechanisms of sparse and reliable visual cortical activity during nonclassical receptive field stimulation[J]. Neuron ,2010,65: 107–121.
[7] Foldiak P. Forming sparse representations by a local anti-hebbian rule[J]. Biol Cybern,1990, 64: 165–170.
[8] Tolhurst DJ, Smyth D, Thompson ID. The sparseness of neuronal responses in ferret primary visual cortex[J].
Neurosci ,2009,29: 2355–2370.
[9] Olshausen BA, Cadieu CF, Warland DK. Learning real and complex overcomplete representations from the statistics of natural images[J]. Proc SPIE,2009,7446: 74460S-1–74460S-11.
[10] Olshausen BA. Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J]. Nature,1996, 381: 607–609.
[11] Bailu S. The role of competitive learning in the generation of DG fields from EC inputs. Cognitive Neurodynamics,2009, 3:177.
[12] Simon M, Ulrich G. Neuromodulation of STDP through short-term changes in firing causality[J]. CognitiveNeurodynamics,2012, 6:353-366.
[13] Dan Y, Poo M. Spike timing-dependent plasticity of neural circuits[J]. Neuron ,2004,44:23–30.
[14] Feldman DE. Synaptic mechanisms for plasticity in neocortex[J]. Annu Rev Neurosci ,2009,32:33–55.
[15] Clopath C, Bu sing L, Gerstner W. Connectivity reflects coding: a model of voltage-based STDP with homeostasis[J]. Nat Neurosci,2010, 13:344–352.
[16] Rehn M, Sommer FT. A network that uses few active neurones to code visual input predicts the diverse shapes of cortical receptive fields[J]. Comput Neurosci ,2007,22: 135–146.
[17] Daqing Guo, Chunguang Li. Population rate coding in recurrent neuronal networks with unreliable synapses[J]. Cognitive Neuro dynamics ,2012,6:75.
[18] Rumelhart, David E, David Zipser. Feature discovery by competitive learning[J]. Cognitivescience,1985,9.1:75-112.
[19] John G N, Robert M, Bruce G W. From Neuron to Brain. Fourth editor. Sinauer Association. 2001.
[20] Haykin, Simon. A comprehensive foundation[J]. Neural Networks,2004,2:2004.
[21] Perrinet LU. Role of homeostasis in learning sparse representations[J]. Neural Comput ,2010,22:1812–1836.
[22] Foldiak P. Forming sparse representations by local anti-Hebbian learning[J]. Biol Cybern ,1990,64:165–170.
[23] Oja E. As implified neuron model as a principal component analyzer[J]. J Math Biol ,1982,15:267–273.
[24] Ringach D. Spatial structure and asymmetry of simple-cell receptive fields in macaque primary visual cortex. J Neurophysiol ,2002,88: 455–463.
[25] Markram H, Toledo-Rodriguez M, Wang Y, Gupta A, Silberberg G, Wu C. Interneurons of the neocortical inhibitory system[J]. Nat Rev Neurosci ,2004,5:793–807.
[26] Rozell CJ, Johnson DH, Baraniuk RG, Olshausen BA. Sparse coding via thresholding and local competition in neural circuits[J].Neural Comput ,2008,20:2526–2563.
[27] Shapero S, Bruderle D, Hasler P, RozellC. Sparse approximation on a network of locally competitive integrate and fire neurons. In Computational and Systems Neuroscience (COSYNE) Meeting, Salt Lake City, UT, February 2011
[28] Hu T, Genkin A, Chklovskii D. Computing sparse representations using a network of integrate-and-fire neurons, J-anelia Farm Research Campus. In Computational and Systems Neuroscien (COSYNE)Meeting, Salt Lake City,
February 27, 2012.
[29] Jones J, Almer L. An evaluation of the two-dimensional Gabor filter model of simple receptive fields in cat stri-ate cortex[J]. J Neurophysiol 58:1233–1258, 1987a.