Deep neural networks: a new framework for modelling biological vision and brain information processing

By Nikolaus Kriegeskorte. 2015

1,125 citations recorded by the source corpus.

Recent advances in neural network modelling have enabled major strides in computer vision and other artificial intelligence applications. Human-level visual recognition abilities are coming within reach of artificial systems. Artificial neural networks are inspired by the brain and their computations could be implemented in biological neurons. Convolutional feedforward networks, which now dominate computer vision, take further inspiration from the architecture of the primate visual hierarchy. However, the current models are designed with engineering goals and not to model brain computations. Nevertheless, initial studies comparing internal representations between these models and primate brains find surprisingly similar representational spaces. With human-level performance no longer out of reach, we are entering an exciting new era, in which we will be able to build neurobiologically faithful feedforward and recurrent computational models of how biological brains perform high-level feats of intelligence, including vision.

Subfields: Artificial Intelligence, Biophysics, Cognitive Neuroscience, Computer Vision and Pattern Recognition, Model Architectures

How researchers see this paper

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

In Artificial Intelligence

In Biophysics

In Cognitive Neuroscience

In Computer Vision and Pattern Recognition

In Model Architectures

Citations in the corpus

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness (2018, background): “As Kriegeskorte (2015) puts it, “the network acquires complex knowledge about the kinds of shapes associated with each category.”
What can the brain teach us about building artificial intelligence? (2016, methodology): “ the high-level feature representations of these convolutional nets have also been used to predict patterns of neural response in human and macaque IT cortex (Khaligh-Razavi & Kriegeskorte, 2014; Kriegeskorte, 2015; Yamins et al., 2014) as well as human typicality ratings (Lake, Zaremba, Fergus,”
Generalisation in humans and deep neural networks (2018, background): “On the other hand, often the strength of a model lies not in replicating the original system but rather in its ability to capture the important aspects while abstracting from details of the implementation (e.g. [26, 27]).”
Biologically inspired protection of deep networks from adversarial attacks (2017, background): “tem. 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). So their susceptibility to adversarial perturbations that are imperceptible to ”
Improving Visual Prompt Tuning for Self-supervised Vision Transformers (2023, background): “It is well-known that Deep Neural Networks (DNNs) progressively abstract the information contained in data samples across their layers (LeCun et al., 2015; Schmidhuber, 2015; Kriegeskorte, 2015).”
Inferring brain-computational mechanisms with models of activity measurements (2016, background): “B 371: 20160278. http://dx.doi.org/10.1098/rstb.2016.0278 Accepted: 19 July 2016 One contribution of 15 to a Theo Murphy meeting issue ‘Interpreting BOLD: a dialogue between cognitive and cellular neuroscience’.”
Building Machines that Learn and Think Like People (2018, methodology): “, 2014), where the high-level feature representations of these convolutional nets have also been used to predict patterns of neural response in human and macaque IT cortex (Khaligh-Razavi & Kriegeskorte, 2014; Kriegeskorte, 2015; Yamins et al., 2014) and human typicality ratings for images of common”
Correspondence of Deep Neural Networks and the Brain for Visual Textures (2018, background): “Although CNNs ([3], [4], [1],[5]) are only crudely matched to the hierarchical structure of the brain, such models have been intriguingly able to predict some aspects of cortical visual processing [6], [7], [8], [9], [10], [11], [12], [13], [14].”
Presynaptic stochasticity improves energy efficiency and helps alleviate the stability-plasticity dilemma (2021, background): “As a growing body of literature suggests, artificial neural networks (ANNs) match several aspects of biological neuronal networks in various goal-driven situations (Kriegeskorte, 2015; Yamins and DiCarlo, 2016; Kell et al., 2018; Banino et al., 2018; Cueva and Wei, 2018; Mattar and Daw, 2018).”
A Developmental Approach to Machine Learning? (2017, background): “There is growing excitement (Cadieu et al., 2014; Kriegeskorte, 2015; Marblestone et al., 2016) that we are at the tipping point for powerful new insights into both human and artiﬁcial intelligence and that these insights will emerge more rapidly by explicitly connecting advances in human…”
Deep Neural Networks as Scientific Models. (2019, background): “Thus, DNNs shape what companies want to invest in, students want to learn, and researchers want to research. phenomena such as human behaviour and brain activity that depend on large amounts of world knowledge that cannot be easily summarised [63].”
Single Subject Prediction of Brain Disorders in Neuroimaging: Promises and Pitfalls (2016, background): “5% Lahmiri et al. (2014) t2:79 sMRI AD/MCI Volumetric measures 120 (115 brain features) LDA CN = 125, AD = 55, EMCI = 114, LMCI = 91, Total = 385 90.”
Synthetic-Neuroscore: Using a neuro-AI interface for evaluating generative adversarial networks (2019, background): “…success achieved by deep neural networks (DNNs) in areas including computer vision and natural language processing, the operation and functionality of DNNs and its connection with the human brain has been extensively studied and investigated in the literature [22, 23, 24, 25, 26, 27, 18, 28, 29].”
Brains and algorithms partially converge in natural language processing (2022, methodology): “This embedding was used to replicate and extend previous work on the similarity between visual neural network activations and brain responses to the same images (e.g., 42,52,53 ).”
Shape or Texture: Understanding Discriminative Features in CNNs (2021, background): “The intuition is that this hierarchical learning of latent representations allows CNNs to recognize complex object shapes to correctly classify images (Kriegeskorte, 2015).”
Cognitive computational neuroscience (2018, background): “Like brains, neural network models can perform feedforward as well as recurrent computations.(50,106) The models driving the recent advances are deep in the sense that they comprise multiple stages of linearnonlinear signal transformation.”