Performance-optimized hierarchical models predict neural responses in higher visual cortex

By Daniel L. K. Yamins, Ha Hong, Charles F. Cadieu, E. Solomon, Darren Seibert, James J. DiCarlo. 2014

2,251 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition, Neuroscience

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 Computer Vision and Pattern Recognition

In Neuroscience

Citations in the corpus

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness (2018, methodology): “Furthermore, CNNs are currently the most predictive models for human ventral stream object recognition (e.g. Cadieu et al., 2014; Yamins et al., 2014); and it is well-known that object shape is the single most important cue for human object recognition (Landau et al.”
What can the brain teach us about building artificial intelligence? (2016, background): “presentations are remarkably similar to those of the human ventral stream (Eickenberg, Gramfort, Varoquaux, & Thirion, 2016; Güçlü & van Gerven, 2015; Khaligh-Razavi & Kriegeskorte, 2014; Yamins et al., 2014). However, even at a glance, we understand visual scenes much more deeply than c”
A Neural Algorithm of Artistic Style (2015, neutral): “image and the image that is generated and Al and Gl their respective style representations in layer l. The contribution of that layer to the total loss is then E l = 1 4N2 l M 2 l X i;j G l ij A ij 2 (4) and the total loss is L style(~a;~x) = XL l=0 w lE l (5) where w l are weighting factors of the ”
Generalisation in humans and deep neural networks (2018, methodology): “Since DNNs and humans achieve similar accuracy, a number of studies have started investigating similarities and differences between DNNs and human vision [9–24].”
Similarity of Neural Network Representations Revisited (2019, background): “, 2016), as well as between brains and neural networks (Yamins et al., 2014; Khaligh-Razavi & Kriegeskorte, 2014; Sussillo et al., 2015).”
Biologically inspired protection of deep networks from adversarial attacks (2017, background): “This discrepancy is of particular scientiﬁc concern as deep neural networks now form foundational models in neuroscience for the visual processing stream (Yamins et al., 2014; McIntosh et al., 2016; Kriegeskorte, 2015).”
Resurrecting the sigmoid in deep learning through dynamical isometry: theory and practice (2017, background): “1 Introduction Deep learning has achieved state-of-the-art performance in many domains, including computer vision [1], machine translation [2], human games [3], education [4], and neurobiological modelling [5, 6].”
Cognitive Science in the era of Artificial Intelligence: A roadmap for reverse-engineering the infant language-learner (2016, positive): “For instance, a deep neural network trained to recognize artefacts and natural kind categories from images turn out to be good predictors of multi-unit responses of neurons in the Inferior Temporal cortex of primates (e.g., Cadieu et al., 2014; Yamins et al., 2014).”
Texture synthesis and the controlled generation of natural stimuli using convolutional neural networks (2015, methodology): “For areas V4 and IT this was previously demonstrated by linearly decoding the identity of presented objects from neural activity, which worked signiﬁcantly better from the neural representation in IT than V4 [21].”
Unsupervised Learning From Video With Deep Neural Embeddings (2019, neutral): “This result is, interestingly, consistent with observations from neuroscience, where it has been shown that both ventral stream [38] and dorsal stream [8] brain pathways contribute to visual performance, with the former being more sensitive to static objects and the”
Synthetic-Neuroscore: Using a neuro-AI interface for evaluating generative adversarial networks (2019, background): “The current literature demonstrates that a CNN is able to predict neural responses in the inferior temporal cortex in an image recognition task [17, 18] via invasive BCI techniques [19].”
Partial success in closing the gap between human and machine vision (2021, neutral): “This includes comparing against neural data [84, 85], contrasting Gestalt effects [e.g. 86], object similarity judgments [87], or mid-level properties [61] and is of course not limited to studying object recognition.”
Brain-Score: Which Artificial Neural Network for Object Recognition is most Brain-Like? (2018, methodology): “• We replicate prior work (Yamins et al., 2014) showing that ANNs that have higher ImageNet performance tend to be more functionally similar to the ventral visual stream, and we extend that work by demonstrating that many state-of-the-art deep ANNs simultaneously score well on all three of the brain”