White Noise Analysis of Neural Networks

A white noise analysis of modern deep neural networks is presented to unveil their biases at the whole network level or the single neuron level. Our analysis is based on two popular and related methods in psychophysics and neurophysiology namely classification images and spike triggered analysis. These methods have been widely used to understand the underlying mechanisms of sensory systems in humans and monkeys. We leverage them to investigate the inherent biases of deep neural networks and to obtain a first-order approximation of their functionality. We emphasize on CNNs since they are currently the state of the art methods in computer vision and are a decent model of human visual processing. In addition, we study multi-layer perceptrons, logistic regression, and recurrent neural networks. Experiments over four classic datasets, MNIST, Fashion-MNIST, CIFAR-10, and ImageNet, show that the computed bias maps resemble the target classes and when used for classification lead to an over twofold performance than the chance level. Further, we show that classification images can be used to attack a black-box classifier and to detect adversarial patch attacks. Finally, we utilize spike triggered averaging to derive the filters of CNNs and explore how the behavior of a network changes when neurons in different layers are modulated. Our effort illustrates a successful example of borrowing from neurosciences to study ANNs and highlights the importance of cross-fertilization and synergy across machine learning, deep learning, and computational neuroscience.

[1]  Il Memming Park,et al.  Bayesian Spike-Triggered Covariance Analysis , 2011, NIPS.

[2]  Eero P. Simoncelli,et al.  Spike-triggered neural characterization. , 2006, Journal of vision.

[3]  Stefano Panzeri,et al.  Inference of neuronal functional circuitry with spike-triggered non-negative matrix factorization , 2017, Nature Communications.

[4]  Anqi Wu,et al.  Convolutional spike-triggered covariance analysis for neural subunit models , 2015, NIPS.

[5]  Jason Yosinski,et al.  Understanding Neural Networks via Feature Visualization: A survey , 2019, Explainable AI.

[6]  I. Ohzawa,et al.  Spatiotemporal organization of simple-cell receptive fields in the cat's striate cortex. II. Linearity of temporal and spatial summation. , 1993, Journal of neurophysiology.

[7]  F A Wichmann,et al.  Ning for Helpful Comments and Suggestions. This Paper Benefited Con- Siderably from Conscientious Peer Review, and We Thank Our Reviewers the Psychometric Function: I. Fitting, Sampling, and Goodness of Fit , 2001 .

[8]  Michelle R. Greene,et al.  Visual Noise from Natural Scene Statistics Reveals Human Scene Category Representations , 2014, ArXiv.

[9]  Jason Yosinski,et al.  Deep neural networks are easily fooled: High confidence predictions for unrecognizable images , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[10]  Pouya Bashivan,et al.  Neural Population Control via Deep Image Synthesis , 2018 .

[11]  Roland Vollgraf,et al.  Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms , 2017, ArXiv.

[12]  J. P. Jones,et al.  The two-dimensional spatial structure of simple receptive fields in cat striate cortex. , 1987, Journal of neurophysiology.

[13]  John Bickle,et al.  Revolutions in Neuroscience: Tool Development , 2016, Front. Syst. Neurosci..

[14]  V. Marmarelis,et al.  Understanding spike-triggered covariance using Wiener theory for receptive field identification. , 2015, Journal of vision.

[15]  Geoffrey E. Hinton,et al.  Matrix capsules with EM routing , 2018, ICLR.

[16]  Fei-Fei Li,et al.  ImageNet: A large-scale hierarchical image database , 2009, 2009 IEEE Conference on Computer Vision and Pattern Recognition.

[17]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[18]  P. Z. Marmarelis,et al.  Analysis of Physiological Systems: The White-Noise Approach , 2011 .

[19]  Benjamin Recht,et al.  Do ImageNet Classifiers Generalize to ImageNet? , 2019, ICML.

[20]  Miguel P Eckstein,et al.  The time course of visual information accrual guiding eye movement decisions. , 2004, Proceedings of the National Academy of Sciences of the United States of America.

[21]  Joan Bruna,et al.  Intriguing properties of neural networks , 2013, ICLR.

[22]  Philip J. Kellman,et al.  Classification images reveal spatiotemporal contour interpolation , 2007, Vision Research.

[23]  Walter J. Scheirer,et al.  Using human brain activity to guide machine learning , 2017, Scientific Reports.

[24]  Rob Fergus,et al.  Visualizing and Understanding Convolutional Networks , 2013, ECCV.

[25]  Jeffrey V Rosenfeld,et al.  Brain Neuromodulation Techniques , 2016, The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry.

[26]  Antonio Torralba,et al.  Learning visual biases from human imagination , 2014, NIPS.

[27]  Sergey Ioffe,et al.  Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning , 2016, AAAI.

[28]  K Naka,et al.  Signal transmission in the catfish retina. I. Transmission in the outer retina. , 1985, Journal of neurophysiology.

[29]  A. Bovik,et al.  Visual search in noise: revealing the influence of structural cues by gaze-contingent classification image analysis. , 2006, Journal of vision.

[30]  W. Newsome,et al.  What electrical microstimulation has revealed about the neural basis of cognition , 2004, Current Opinion in Neurobiology.

[31]  Alex Krizhevsky,et al.  Learning Multiple Layers of Features from Tiny Images , 2009 .

[32]  A. Ahumada.  Perceptual Classification Images from Vernier Acuity Masked by Noise , 1996 .

[33]  Ali Borji,et al.  State-of-the-Art in Visual Attention Modeling , 2013, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[34]  Andrea Vedaldi,et al.  Understanding deep image representations by inverting them , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[35]  Pascal Vincent,et al.  Visualizing Higher-Layer Features of a Deep Network , 2009 .

[36]  R. Kiani,et al.  Microstimulation of inferotemporal cortex influences face categorization , 2006, Nature.

[37]  Markus Meister,et al.  Multi-neuronal signals from the retina: acquisition and analysis , 1994, Journal of Neuroscience Methods.

[38]  Jonathon Shlens,et al.  Explaining and Harnessing Adversarial Examples , 2014, ICLR.

[39]  T. Moore,et al.  Microstimulation of the frontal eye field and its effects on covert spatial attention. , 2004, Journal of neurophysiology.

[40]  Bolei Zhou,et al.  Network Dissection: Quantifying Interpretability of Deep Visual Representations , 2017, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[41]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[42]  Richard F Murray,et al.  Classification images: A review. , 2011, Journal of vision.

[43]  Antonio Torralba,et al.  HOGgles: Visualizing Object Detection Features , 2013, 2013 IEEE International Conference on Computer Vision.

[44]  R. Reid,et al.  Specificity of monosynaptic connections from thalamus to visual cortex , 1995, Nature.

[45]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[46]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[47]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[48]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.