Hierarchical Modular Optimization of Convolutional Networks Achieves Representations Similar to Macaque IT and Human Ventral Stream

By Daniel L. K. Yamins, Ha Hong, Charles F. Cadieu, James J. DiCarlo. 2013

171 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Active Long Term Memory Networks (2016, background): “We conjecture that CI arises because during the learning lifespan of a system the distribution of locally observed states [22, 23] in the environment is non-stationary and potentially chaotic, while the neural representation of the environment with respect to its mode of variation as latent graphica”
Deep Neural Networks Reveal a Gradient in the Complexity of Neural Representations across the Ventral Stream (2014, positive): “llow decoding of stimuli from observed neural responses [16]. Recently, it was shown that performance-optimized hierarchical models can predict singleneuron responses in area IT of the macaque monkey [17]. Our current work signicantly expands on these important results in (i) showing that there is a”
Using goal-driven deep learning models to understand sensory cortex (2016, background): “Though parameter choices might seem like mere details, subtle parameter differences can dramatically affect a network’s performance on recognition tasks and its match to neural data 15,18 .”
A parametric texture model based on deep convolutional features closely matches texture appearance for humans (2017, unknown): “…models built explicitly for that purpose ( Cadieu et al., 2014; Cichy, Khosla, Pantazis, & Oliva, 2016; Cichy, Khosla, Pantazis, Torralba, & Oliva, 2016; Guclu & van Gerven, 2015; Hong et al., 2016; Khaligh-Razavi & Kriegeskorte , 2014; Yamins, Hong, Cadieu, & DiCarlo, 2013; Yamins et al., 2014).”
Performance-optimized hierarchical models predict neural responses in higher visual cortex (2014, methodology): “The HMO procedure (31) is a computational optimization procedure designed to identify high-performing network architectures from the space N.”
Neural population control via deep image synthesis (2018, background): “…to the visually evoked neural representations in mid-level (area V4) and high-level (inferior temporal) cortical stages of the ventral stream ( 2 , 3 ) — a finding that has been extended to neural re-presentations invisualareaV1 ( 4 ), topatterns of behavioral performance in core object…”
Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition (2014, methodology): “Our methodology of analyzing absolute representational performance using kernel analysis provides a novel and complementary finding to the results in [27,32,34].”
Unsupervised learning of invariant representations with low sample complexity: the magic of sensory cortex or a new framework for machine learning? (2013, background): “One of the observations that inspired our theory is that in convolutional architectures, random features can often perform nearly as well as features learned from objects [61, 62, 13, 59] – the architecture often matters more than the particular features computed.”
Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs (2019, methodology): “Surprisingly, after training deep, feedforward ANNs to perform the standard ImageNet categorization task [5], intermediate layers in ANNs can partly account for how neurons in intermediate layers of the primate visual system will respond to any given image, even ones that the model has never seen be”
Deep neural networks: a new framework for modelling biological vision and brain information processing (2015, background): “large set of automatically generated neural net architectures using random features (Yamins et al. 2013; 2014), for a wide range of popular hand-engineered computer vision features and neuroscientific vision models (Khaligh-Razavi et al.”
Deep Supervised, but Not Unsupervised, Models May Explain IT Cortical Representation (2014, methodology): “The analysis of our human and monkey data in [77] similarly found that a hierarchical feedforward model optimized for invariant object recognition could account for the IT representational geometry only when linear readout features emphasizing the appropriate categorical divisions were fitted to the”