Colloquium Cognitive Systems

Connecting Cognition to the Brain: a Spiking-Neuron model of Associative Recognition

Jelmer Borst (University of Groningen, NL)

Abstract:  Most cognitive and mathematical models describe the human mind at the high level: for example, cognitive architectures like ACT-R use concepts like ‘memory retrievals’ and ‘chunks’, evidence accumulation models focus on the decision process, and Bayesian models try to understand why the human mind functions as it does. However, if we want to understand how cognition emerges from the human brain, we need to develop cognitive brain models. In recent years, several neural models have been proposed. Often, these models are tremendously detailed (e.g., Markram's Blue Brain Project), but unfortunately fail to make the connection to actual behaviour, or only focus on qualitative patterns. In addition, even if neural models are linked to functional behaviour (e.g. Eliasmith's Spaun), such models are necessarily so complex that they are very hard to evaluate. To overcome these problems, we developed an integrative biologically-plausible brain model of associative recognition that matches human behavior quantitively. To further evaluate the resulting model, we used its neural spikes to predict source-localized MEG and fMRI data. The results matched source-localized MEG data in occipital, temporal, prefrontal, and precentral brain regions; as well as a classic fMRI effect in prefrontal cortex. Integrating across different levels of description - from neurons to cognition and from spikes to MEG/fMRI data -  resulted in a model that can explain how the human mind might be implemented by the brain. 

About:
Jelmer Borst is an associate professor in computational cognitve neuroscience at the University of Groningen in the Netherlands. He is an expert on using machine-learning methods to combine psychophysiological data (e.g., pupil dilation) and neuroimaging data (fMRI, MEG, EEG) with computational cognitive models (e.g. ACT-R, spiking-neural networks). He was one of the core developers of the HMP method, which yields unparalleled insight into stages of cognitive processing (github.com/GWeindel/hmp). His research is funded by the Netherlands Organization of Scientific Research, The European Office of Aerospace Research and Development, and internal university grants.