Rufin VanRullen

research topics

Our group works at the interface of AI and Neuroscience.
Our research aims are to:

• Design novel state-of-the-art AI systems and deep learning algorithms, by drawing on knowledge from cognitive neuroscience.

  For instance:
  • Latent Global Workspace models for flexible cognition (see ERC GLoW)
    VanRullen & Kanai (Trends in Neurosciences, 2021)
  • Recurrent neural networks based on the predictive coding framework
    Choksi, Mozafari, Biggs O'May, Ador, Alamia & VanRullen (NeurIPS 2021)
  • Global attention for convolutional networks
    VanRullen & Alamia (ICANN-International Conf. on Artificial Neural Networks, 2021)

• Understand brain function, by developing large-scale computational models and neural networks.

  For instance:
  • Human-like illusion perception in predictive feedback neural networks
    Pang, Biggs O'May, Choksi & VanRullen (Neural Networks, 2021)
  • Emergence of oscillations and travelling waves in predictive feedback networks
    Alamia & VanRullen (PLoS Biology, 2019)

• More generally, bridge the gap between biological and machine intelligence.

  For instance:
  • Comparing brains and deep nets in visual reasoning
    Alamia, Luo, Ricci, Kim, Serre & VanRullen (eNeuro, 2020)
  • Decoding brain activity with deep learning
    VanRullen & Reddy (Communications Biology, 2019)
    Mozafari, Reddy & VanRullen (IJCNN-International Joint Conf. on Neural Networks, 2020)
    

Previously...

Our research used to have a stronger experimental neuroscience component. We studied the functional consequences of brain oscillations - in particular, how:

• Brain oscillations discretize sensation into "perceptual cycles"
  VanRullen (Trends in Cogn. Sciences 2016)
• Attention operates rhythmically, i.e. "attention cycles"
  VanRullen (Neuron, 2018)

For more details