Comparison of neuronal responses in primate inferior-temporal cortex and feed-forward deep neural network model with regard to information processing of faces

Matsumoto, Narihisa; Mototake, Yoh-ichi; Kawano, Kenji; Okada, Masato; Sugase-Miyamoto, Yasuko

doi:10.1007/s10827-021-00778-5

Cited by 4 publications

(5 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The model we constructed for that study used random bit patterns not visual images as input. In another study, we constructed a DNN, i.e., AlexNet ( Krizhevsky et al, 2012 ), to compare the information represented in each layer and the information encoded by a neural population in area TE with a visual stimulus set that included human and monkey faces ( Matsumoto et al, 2021 ). Thus, the representation in the fully connected layers of AlexNet most resembled the representation of TE neurons for human and monkey faces.…”

Section: Discussionmentioning

confidence: 99%

“…The fully connected layer was removed from the original Xception net, and the Hopfield model was inserted instead as a model of area TE. This was done because our previous studies showed that the information representation in fully connected layers of a DNN was similar to the representation in area TE ( Matsumoto et al, 2021 ) and that an associative memory model was able to reproduce the neural activities of area TE ( Matsumoto et al, 2005b ). We compared the performance of the combined model with another model, i.e., the Xception model without the Hopfield model ( Figure 1B ).…”

Section: Methodsmentioning

confidence: 99%

“…In area TE, some neurons respond to complex objects, faces, and so on and represent information about a global category, e.g., human vs. monkey vs. simple shapes, earlier than fine category information about faces, e.g., facial expression or identity ( Sugase et al, 1999 ; Matsumoto et al, 2005a ; Sugase-Miyamoto et al, 2014 ). In our previous study, we constructed a deep neural network (DNN) to compare information representation in each layer and information encoded by a neural population in area TE with a visual stimulus set that included human and monkey faces ( Matsumoto et al, 2021 ). We found that the time course of hierarchical categorization could not be replicated with the DNN.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Recurrent Connections Might Be Important for Hierarchical Categorization

Matsumoto

Taguchi

Shimizu

et al. 2022

Front. Syst. Neurosci.

Self Cite

View full text Add to dashboard Cite

Visual short-term memory is an important ability of primates and is thought to be stored in area TE. We previously reported that the initial transient responses of neurons in area TE represented information about a global category of faces, e.g., monkey faces vs. human faces vs. simple shapes, and the latter part of the responses represented information about fine categories, e.g., facial expression. The neuronal mechanisms of hierarchical categorization in area TE remain unknown. For this study, we constructed a combined model that consisted of a deep neural network (DNN) and a recurrent neural network and investigated whether this model can replicate the time course of hierarchical categorization. The visual images were stored in the recurrent connections of the model. When the visual images with noise were input to the model, the model outputted the time course of the hierarchical categorization. This result indicates that recurrent connections in the model are important not only for visual short-term memory but for hierarchical categorization, suggesting that recurrent connections in area TE are important for hierarchical categorization.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Recurrent Connections Might Be Important for Hierarchical Categorization

Matsumoto

Taguchi

Shimizu

et al. 2022

Front. Syst. Neurosci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The outputs from mid layers of deep neural networks, trained for object recognition tasks, are able to model spiking activities in area V1 of monkeys, and Deep Neural Network (DNN) models have been able to predict spiking activity in V1 area of monkeys better than the previous models [ 31 ]. Earlier studies have indicated correlations between the output of hidden layers of a trained Alexnet and brain electrode readings from monkeys presented with the same stimulus [ 32 ]. The outputs of intermediate layers of a DNN trained on object classification, gives responses similar to neural responses of inferior temporal cortex and area V4 of rhesus macaques presented with the same task.…”

Section: Related Workmentioning

confidence: 99%

A deep learning based cognitive model to probe the relation between psychophysics and electrophysiology of flicker stimulus

Chandran,

Ghosh

2024

Brain Inf.

View full text Add to dashboard Cite

The flicker stimulus is a visual stimulus of intermittent illumination. A flicker stimulus can appear flickering or steady to a human subject, depending on the physical parameters associated with the stimulus. When the flickering light appears steady, flicker fusion is said to have occurred. This work aims to bridge the gap between the psychophysics of flicker fusion and the electrophysiology associated with flicker stimulus through a Deep Learning based computational model of flicker perception. Convolutional Recurrent Neural Networks (CRNNs) were trained with psychophysics data of flicker stimulus obtained from a human subject. We claim that many of the reported features of electrophysiology of the flicker stimulus, including the presence of fundamentals and harmonics of the stimulus, can be explained as the result of a temporal convolution operation on the flicker stimulus. We further show that the convolution layer output of a CRNN trained with psychophysics data is more responsive to specific frequencies as in human EEG response to flicker, and the convolution layer of a trained CRNN can give a nearly sinusoidal output for 10 hertz flicker stimulus as reported for some human subjects.

show abstract

“…The network architecture is designed to coordinate and optimize distributed system control processes (as shown in Figure 1) to meet the needs of users, which is also the issue we need to study [9][10]. In this article, the network groups include:…”

Section: Figure1 Network Architecturementioning

confidence: 99%

Distributed System Coordination Predictive Control for Network Information Mode

2022

DPS

View full text Add to dashboard Cite

With the rapid development of network information technology, distributed system coordinated control has become an indispensable part of the network information industry in the world today. In this paper, the concepts and classifications of the distributed system coordinated control method for network information mode are introduced first, then the problems existing in the traditional network structure are analyzed in detail. Then, from the point of view of genetic algorithm theory, a support vector machine (SVM), fuzzy logic and transfer function are established to solve the actual non-linear load dispatch, in which the data is processed by neural calculation. Finally, the improved decision table achieves the same control effect as the original model through experimental validation, and improves the system coordination performance index and stability. The experimental results show that the prediction time of this model is within the range of 4s-7s, and its stability, feasibility and prediction accuracy are up to 90%. This indicates that the distributed system coordination prediction model under network information mode performs well.

show abstract

Comparison of neuronal responses in primate inferior-temporal cortex and feed-forward deep neural network model with regard to information processing of faces

Cited by 4 publications

References 18 publications

Recurrent Connections Might Be Important for Hierarchical Categorization

Recurrent Connections Might Be Important for Hierarchical Categorization

A deep learning based cognitive model to probe the relation between psychophysics and electrophysiology of flicker stimulus

Distributed System Coordination Predictive Control for Network Information Mode

Contact Info

Product

Resources

About