Recurrent convolutional neural networks: a better model of biological object recognition

Spoerer, Courtney J.; McClure, Patrick; Kriegeskorte, Nikolaus

doi:10.1101/133330

Cited by 73 publications

(111 citation statements)

References 50 publications

Supporting

Mentioning

104

Contrasting

Order By: Relevance

“…Together with the large scale iEEG recordings of human visual columns they allowed fresh insights into the functional role and mechanistic generation of high order human face representations. Employing the rapidly evolving new DCNNs may help, in the future, to resolve outstanding issues such as the functionalities of distinct cortical patches in high order visual areas and the functional role of top down and local recurrent processing in brain function (Spoerer, McClure, & Kriegeskorte, 2017).…”

Section: Discussionmentioning

confidence: 99%

Deep Convolutional modeling of human face selective columns reveals their role in pictorial face representation

Grossman

Gaziv

et al. 2018

Preprint

View full text Add to dashboard Cite

Despite the massive accumulation of systems neuroscience findings, their functional meaning remains tentative, largely due to the absence of realistically performing models. The discovery that deep convolutional networks achieve human performance in realistic tasks offers fresh opportunities for such modeling. Here we show that the face-space topography of face-selective columns recorded intracranially in 32 patients significantly matches that of a DCNN having human-level face recognition capabilities. Three modeling aspects converge in pointing to a role of human face areas in pictorial rather than person identification: First, the match was confined to intermediate layers of the DCNN. Second, identity preserving image manipulations abolished the brain to DCNN correlation. Third, DCNN neurons matching face-column tuning displayed view-point selective receptive fields. Our results point to a "convergent evolution" of pattern similarities in biological and artificial face perception. They demonstrate DCNNs as a powerful modeling approach for deciphering the function of human cortical networks.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep Convolutional modeling of human face selective columns reveals their role in pictorial face representation

Grossman

Gaziv

et al. 2018

Preprint

View full text Add to dashboard Cite

show abstract

“…The timing of signatures of facial identity (Barragan-Jason, Besson, Ceccaldi, & Barbeau, 2013;Freiwald & Tsao, 2010) and social cues, such as direct eye-contact (Kietzmann et al, 2017), too, point towards a reliance on recurrent computations. Finally, recurrent connections likely play a vital role in early category learning (Kietzmann, Ehinger, Porada, Engel, & König, 2016), in dealing with occlusion (Oord, Kalchbrenner, & Kavukcuoglu, 2016;Spoerer, McClure, & Kriegeskorte, 2017;Wyatte, Curran, & O'Reilly, 2012;Wyatte, Jilk, & O'Reilly, 2014) and object-based attention (Roelfsema, Lamme, & Spekreijse, 1998).…”

Section: Beyond the Feed-forward Sweep: Recurrent Dnnsmentioning

confidence: 99%

Deep Neural Networks in Computational Neuroscience

Kietzmann

McClure

Kriegeskorte

2017

Preprint

Self Cite

View full text Add to dashboard Cite

Keywords: deep neural networks, deep learning, convolutional neural networks, objective functions, recurrence, black box, levels of abstraction, modelling the brain, input statistics, biological detail SummaryThe goal of computational neuroscience is to find mechanistic explanations of how the nervous system processes information to give rise to cognitive function and behaviour. At the heart of the field are its models, i.e. mathematical and computational descriptions of the system being studied, which map sensory stimuli to neural responses and/or neural to behavioural responses. These models range from simple to complex. Recently, deep neural networks (DNNs) have come to dominate several domains of artificial intelligence (AI). As the term "neural network" suggests, these models are inspired by biological brains. However, current DNNs neglect many details of biological neural networks. These simplifications contribute to their computational efficiency, enabling them to perform complex feats of intelligence, ranging from perceptual (e.g. visual object and auditory speech recognition) to cognitive tasks (e.g. machine translation), and on to motor control (e.g. playing computer games or controlling a robot arm). In addition to their ability to model complex intelligent behaviours, DNNs excel at predicting neural responses to novel sensory stimuli with accuracies well beyond any other currently available model type. DNNs can have millions of parameters, which are required to capture the domain knowledge needed for successful task performance. Contrary to the intuition that this renders them into impenetrable black boxes, the computational properties of the network units are the result of four directly manipulable elements: input statistics, network structure, functional objective, and learning algorithm. With full access to the activity and connectivity of all units, advanced visualization techniques, and analytic tools to map network representations to neural data, DNNs represent a powerful framework for building task-performing models and will drive substantial insights in computational neuroscience.

show abstract

“…Unlike other studies that use stimuli that are occluded or camouflaged (Spoerer et al, 2017;Tang and Kreiman, 2017), our RSVP task offers no obvious computation that can be embued to feedback processes when presentation times are shortened. That is, our study does not inform on the precise nature of computations needed for stimulus evidence accumulation when presentation times are extremely short.…”

Section: Statiotemporal Bounds For Computational Models Of Visionmentioning

confidence: 99%

Ultra-Rapid Serial Visual Presentation Reveals Dynamics of Feedforward and Feedback Processes in the Ventral Visual Pathway

Mohsenzadeh

Qin

Cichy

et al. 2018

Preprint

View full text Add to dashboard Cite

Human visual recognition activates a dense network of overlapping feedforward and recurrent neuronal processes, making it hard to disentangle processing in the feedforward from the feedback direction. Here, we used ultra-rapid serial visual presentation to suppress sustained activity that blurs the boundaries of processing steps, enabling us to resolve two distinct stages of processing with MEG multivariate pattern classification. The first processing stage was the rapid activation cascade of the bottom-up sweep, which terminated early as visual stimuli were presented at progressively faster rates. The second stage was the emergence of categorical information with peak latency that shifted later in time with progressively faster stimulus presentations, indexing time-consuming recurrent processing. Using MEG-fMRI fusion with representational similarity, we localized recurrent signals in early visual cortex. Together, our findings segregated an initial bottom-up sweep from subsequent feedback processing, and revealed the neural signature of increased recurrent processing demands for challenging viewing conditions.

show abstract

Recurrent convolutional neural networks: a better model of biological object recognition

Abstract: 7Feedforward neural networks provide the dominant model of how the brain performs visual 8 object recognition. However, these networks lack the lateral and feedback connections,

Cited by 73 publications

References 50 publications

Deep Convolutional modeling of human face selective columns reveals their role in pictorial face representation

Deep Convolutional modeling of human face selective columns reveals their role in pictorial face representation

Deep Neural Networks in Computational Neuroscience

Ultra-Rapid Serial Visual Presentation Reveals Dynamics of Feedforward and Feedback Processes in the Ventral Visual Pathway

Contact Info

Product

Resources

About