A feedforward architecture accounts for rapid categorization

By Thomas Serre, Aude Oliva, Tomaso A. Poggio. 2007

1,067 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Molecular Biology

Citations in the corpus

Emergence of Complex-Like Cells in a Temporal Product Network with Local Receptive Fields (2010, background): “There are several system of this class (Lazebnik et al., 2006; Pinto et al., 2008, 2009; Serre et al., 2007; Yang et al., 2009; Lee et al., 2009; Yu et al., 2009) but speciﬁcally convolutional nets achieve 67% (Jarrett et al., 2009).”
Residual Networks are Exponential Ensembles of Relatively Shallow Networks (2016, methodology): “This is contrary to architectures like AlexNet [13] and even biological systems [17] that can capture complex concepts within half a dozen layers.”
Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex (2016, background): “The evolutionary advantage of having fewer layers is apparent: it supports rapid (100msec from image onset to meaningful information in IT neural population) visual recognition, which is a key ability of human and non-human primates [31, 28].”
Toward an Integration of Deep Learning and Neuroscience (2016, methodology): “…noted, though, that recent architectural advances, including specialized spatial attention and feedback mechanisms (Rezende et al., 2016), as well as specialized memory mechanisms (Santoro et al., 2016), do allow some types of one-shot generalization to be driven by backpropagation-based learning.”
The dynamics of invariant object recognition in the human visual system. (2014, methodology): “HMAX performance had a similar trend to the order of the MEG experimental results: an early visual model could not decode stimuli invariant to size or position with above-chance accuracy, a midlevel visual model could decode small transformations with above-chance accuracy, and an IT-like model coul”
A common, high-dimensional model of the representational space in human ventral temporal cortex (2011, background): “Single-neuron response-tuning profiles in monkey inferior temporal cortex (IT) reflect complex object features, and patterns of responses over a population represent object categories and identities (Logothetis and Sheinberg, 1996; Tanaka, 2003; Hung et al., 2005; Tsao et al., 2006; Freiwald et al.,”
Perceived Shape Similarity among Unfamiliar Objects and the Organization of the Human Object Vision Pathway (2008, background): “A more comprehensive computational model of shape recognition might include an adaptive feature dictionary (Serre et al., 2007) so that the tuning becomes more informative for object categorization (Biederman, 1987; Edelman and Intrator, 2000; Ullman et al.”
A Comparative Evaluation of Approximate Probabilistic Simulation and Deep Neural Networks as Accounts of Human Physical Scene Understanding (2016, methodology): “Motivated by the success of CNNs in machine vision object recognition tasks (Krizhevsky et al., 2012), neuroscientists have proposed analogous architectures as accounts of the fast feedforward aspects of human visual object recognition (Yamins et al., 2014; Serre, Oliva, & Poggio, 2007).”
A quantitative theory of immediate visual recognition. (2007, methodology): “We found that the model and human observers exhibited a similar degree of robustness (see Serre et al., 2007a).”
STDP-based spiking deep neural networks for object recognition (2016, background): “We compared the proposed SDNN to several other networks including unsupervised methods such as HMAX and convolutional autoencoder network, and supervised methods such as DCNNs.”
How Deep is the Feature Analysis underlying Rapid Visual Categorization? (2016, methodology): “Some of the ILSVRC best performing architectures now include 150 layers of processing [He et al., 2015] and even 1,000 layers for other recognition challenges [He et al., 2016] – arguably orders of magnitude more than the visual system (estimated to be O (10) [Serre et al., 2007]).”
Interpreting individual classifications of hierarchical networks (2013, background): “In order to make our analysis concrete, in this work we focus on the well-known family of hierarchical networks described by HMAX.”