Explicit information for category-orthogonal object properties increases along the ventral stream

By Ha Hong, Daniel L. K. Yamins, Najib J. Majaj, James J. DiCarlo. Nature Neuroscience, 2016

331 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/09eb087fc5730589e59673aafa0dfc788768611a

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

Citations in the corpus

Deep Neural Networks as Scientific Models. (2019, methodology): “In the ﬁ rst example [35], the authors explored how a DNN trained on object categorisation predicts different object properties.”
Towards the neural population doctrine. (2019, background): “Recently, the authors have extended their results to train a 6 layer CNN to successfully predict category-orthogonal object properties along the ventral stream [55].”
A parametric texture model based on deep convolutional features closely matches texture appearance for humans (2017, methodology): “CNN models also show similarities to human psychophysical object recognition performance under brief presentation conditions (Hong et al., 2016; Yamins et al., 2014).”
Neuroscience-Inspired Artificial Intelligence. (2017, background): “…deep convolutional network architectures offer a computational account of recent neurophysiological data demonstrating that the coding of category-orthogonal properties of objects (e.g., position, size) actually increases as one progresses higher up the ventral visual stream (Hong et al., 2016).”
Representational models: A common framework for understanding encoding, pattern-component, and representational-similarity analysis (2017, background): “Inferior temporal cortex contains a representation of object category [8, 9], along with representations of much additional information [10].”
Cognitive computational neuroscience (2018, background): “Higher layers of deep convolutional neural networks also resemble the IT cortical representation in that both enable the decoding of object position, size, and pose, along with the category of the object.(125) In addition to testing these models by predicting brain-activity data, the field has begun”
What shapes feature representations? Exploring datasets, architectures, and training (2020, background): “rtz and colleagues [38] suggests that deep models compress away task-irrelevant information as they learn a task, though the generality of this observation has been disputed [34]. Several papers have [14, 16] shown that it is possible to decode “categoryorthogonal” features from CNNs trained on natu”
Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future (2020, background): “Other studies have explored tuning to shape and categories [23], and how responses change with changes in pose, location, and size [24, 25, 26].”
Invariant object recognition is a personalized selection of invariant features in humans, not simply explained by hierarchical feed-forward vision models (2017, methodology): “These similarity-dependent results imply that the level of invariance reported in many human psychophysical studies 4,6,48,49 , which has always been a point of conflict 11,54 , could have been influenced by the similarity of objects used in the discrimination task.”
Interpreting neural computations by examining intrinsic and embedding dimensionality of neural activity (2021, background): “For example, assuming that a core function of the early visual system is to categorize objects, modern models of vision are built to make object categories linearly decodable [69].”
Learning Physical Graph Representations from Visual Scenes (2020, background): “This type of information is not easily read out from the raw pixels of RGB movies; but it is known to be decodable from the later, nonlinear layers of task-optimized convolutional neural networks (CNNs) [19].”
Reward-based training of recurrent neural networks for cognitive and value-based tasks (2016, methodology): “In particular, we provide a careful derivation of Equation 3 following, in part, the exposition in Zaremba & Sutskever (2016).”
Large-Scale, High-Resolution Comparison of the Core Visual Object Recognition Behavior of Humans, Monkeys, and State-of-the-Art Deep Artificial Neural Networks (2018, background): “…activity at multiple stages of the primate ventral visual stream (Yamins et al., 2013; Cadieu et al., 2014; Khaligh-Razavi and Kriegeskorte, 2014; Yamins et al., 2014; Güçlü and van Gerven, 2015; Cichy et al., 2016; Hong et al., 2016; Seibert et al., 2016; Cadena et al., 2017; Wen et al., 2017).”