Computation - Theory
Computation - Theory - Machine Learning

Vision is a highly demanding computational problem. Biological and artificial visual systems face the same challenges – e.g., building transformation-invariant representations of visual objects or pattern-invariant representations of motion direction. In our work, we use computational modeling and machine learning tools, often in collaboration with leading theoretical neuroscientists and data scientists, to compare visual processing in brains in machines. Specifically:

  • we have developed conceptual (e.g., ideal observer) models to explain rat visual perception of shapes (Djurdjevic et al., 2018), motion direction (Matteucci et al., 2021) and textures (Caramellino et al., 2021)
  • we have applied cutting edge machine learning approaches (e.g., clustering algorithms) to investigate object representations in monkey (Baldassi et al., 2013) and rat (Vascon et al., 2019) visual cortex
  • we have investigated how data representations are reformatted across deep convolutional neural networks for image classification (Ansuini et al., 2019) and how consistent such representations are with those found along the rat visual cortical hierarchy (Matteucci et al., 2019, 2023; Muratore et al., 2022) and those supporting rat object vision (Muratore et al, 2024).

See below for a complete list of our computational studies.

Selected articles
Unraveling the complexity of rat object vision requires a full convolutional network-and beyond
Muratore P, Alemi A, Zoccolan D (2024)
Biorxiv: doi:
Truly pattern: Nonlinear integration of motion signals is required to account for the responses of pattern cells in rat visual cortex
Matteucci G, Bellacosa Marotti R, Zattera B, Zoccolan D (2023)
Science Adv.: 9 (45), eadh4690
Prune and distill: similar reformatting of image information along rat visual cortex and deep neural networks
Muratore P, Tafazoli S, Piasini E, Laio A, Zoccolan D (2022)
Adv. Neural Info. Processing Systems (NeurIPS): 35
Rat sensitivity to multipoint statistics is predicted by efficient coding of natural scenes
Caramellino R*, Piasini E*, Buccellato A, Carboncino A, Balasubramanian V & Zoccolan D (2021)
eLife: 2021; 10:e72081
Rats spontaneously perceive global motion direction of drifting plaids
Matteucci G*, Zattera B*, Bellacosa Marotti R, & Zoccolan D (2021)
Plos Comp. Biol.: 17(9): e1009415
A machine learning framework to optimize optic nerve electrical stimulation for vision restoration
Romeni S, Zoccolan D & Micera S (2021)
Patterns: 2(7), 100286
Characterization of visual object representations in rat primary visual cortex.
Vascon S*, Parin Y*, Annavini E*, D’Andola M, Zoccolan D & Pelillo M (2019)
ECCV 2018, Lect. Notes Comp. Science: 11131, 577-586
Intrinsic dimension of data representations in deep neural networks.
Ansuini A, Laio A, Macke J & Zoccolan D (2019)
Adv. Neural Info. Processing Systems (NeurIPS): 32
Accuracy of rats in discriminating visual objects is explained by the complexity of their perceptual strategy.
Djurdjevic V*, Ansuini A*, Bertolini D, Macke JH & Zoccolan D (2018)
Curr. Biol. : 28(7), 1005-1015
Shape similarity, better than semantic membership, accounts for the structure of visual object representations in a population of monkey inferotemporal neurons.
Baldassi C*, Alemi-Neissi A*, Pagan M*, DiCarlo JJ, Zecchina R & Zoccolan D (2013)
PLoS Comput. Biol.: 9(8): e1003167
What response properties do individual neurons need to underlie object recognition in clutter?
Li N, Cox DD, Zoccolan D & DiCarlo JJ (2009)
J. Neurophys. : 102, 360-376