Welcome to the page of the Kietzmann Lab, which is part of the Institute of Cognitive Science, University of Osnabrรผck. We investigate principles of neural information processing using tools from machine learning and deep learning, applied to neuroimaging data recorded at high temporal (EEG/MEG) and spatial (fMRI) resolution. Please contact us with any questions or paper requests, and follow Dr. Kietzmann on twitter (@TimKietzmann) for latest lab updates.

Please reach out to us if you are interested in joining the lab and see our page on equity, diversity, and inclusion for further information.

Research Interests

Cognitive Neuroscience meets Machine Learning. Our research group aims to understand the computational processes by which the brain and artificial agents can efficiently and robustly derive meaning from the world around us. We ask how the brain acquires versatile representations from the statistical regularities in the input, how sensory information is dynamically transformed in the cortical network, and which information is extracted by the brain to support higher-level cognition. To find answers to these questions, we develop and employ machine learning techniques to discover and model structure in high-dimensional neural data.

As a target modality, we focus on vision, the most dominant of our senses both neurally and perceptually. To gain insight into the intricate system that enables us to see, the group advances along two interconnected lines of research: machine learning for discovery in neuroscience, and deep neural network modelling. This interdisciplinary work combines machine learning, computational neuroscience, computer vision, and semantics. Our work is therefore at the heart of the emerging fields of neuro-inspired machine learning and cognitive computational neuroscience.

Twitter Feed

By popular demand, we are launching an entire line of "X for Ukraine" t-shirts, where X = Anthropologists, Demographers, etc.

T-shirts were designed in Kyiv, Ukraine, all profits going to humanitarian aid to Ukraine. There's also an option to make an extra donation at checkout.

1/8 Can we use embodied AI to gain insight into *why* neural systems are as they are?

In previous work๐Ÿ‘‡, we demonstrated that a contrastive unsupervised objective substantially outperforms supervised object categorization at generating networks that predict mouse visual cortex. https://twitter.com/aran_nayebi/status/1405883891945291781

Aran Nayebi@aran_nayebi

Here we develop "Unsupervised Models of Mouse Visual Cortex"
Co-lead with @NathanKong
w/ @ChengxuZhuang, Justin Gardner, @amnorcia, @dyamins

#tweetprint below ๐Ÿ‘‡ https://twitter.com/biorxiv_neursci/status/1405665380702601218

The fishing term "bycatch" (i.e. the unintentional catching of non-targeted sea-creatures) works remarkably well for deep nets.

Larger networks will yield more bycatch. In ML: dataset specifics/textures/biases/shortcut learning, not only the signal you are looking for.

New preprint from my PhD (yes I know, 2.5 years later...)! We disentangled information sampling and processing in the brain, observed that information is rhythmically sampled across stimulus time in early visual areas, and more! ๐Ÿงต๐Ÿ‘‡ [1/9] https://bit.ly/3OZhmtC

Load More...