Brain-machine interfaces
Computational Neuroscience Journal Club in the BLG, September 28 2021
Ever since the first recordings of human brain activity around a century ago, developments in recording techniques have greatly improved the quality, variety and scale of neural data collected in neuroscience experiments. To test and expand our understanding of the brain, one has to perform experiments that involve recording and potentially manipulating activity in live neural populations simultaneously with behaviour and/or sensory input. The relation between the observed activity and the external behaviour or input allows one to characterize the meaning and function of such neural signals. Due to the complexity of neural activity, the analysis of neural data relies on powerful statistical tools and computational methods that run on silicon hardware. When combining all these aspects into a single framework, one naturally arrives at the idea of brain-machine interfaces (BMIs) or brain-computer interfaces (BCIs). Such setups aim to provide, as the name suggests, a direct interface to observe and potentially control neural activity through software. If successful, such constructions open up many new avenues for testing neuroscience theories, exploring neural activity, and more practical applications like reading intended behaviour.

Optimal spike coding
Computational Neuroscience Journal Club in the BLG, February 23 2021
An interesting line of research that separate from the conventional rate coding perspective is explored in this talk. This framework is referred to as “optimal spike coding”, named because each individual spike conveys information optimally according to a simple criterion. Interestingly, this framework allows one to “derive” leaky integrate-and-fire dynamics from a normative model, connecting higher level computational functions directly with aspects of single neuron dynamics!

Non-spatial representations in the hippocampal formation
Computational Neuroscience Journal Club in the BLG, October 6 2020
This talk discussed experimental and theoretical works on neural representations in the hippocampal formation that go beyond spatial navigation. Typical cells like place cells exhibit characteristic tuning to animal position, but recent works have found equally informative representations in non-spatial contexts, such as auditory tasks. Theoretical studies hypothesize a general inference framework implemented by such neural representations.

Variability in neural discharge of place and grid cells
Computational Neuroscience Journal Club in the BLG, April 21 2020
This talk discussed experimental and computational analysis of the excess irregularity in neural spiking activity seen in many cells in the hippocampus and entorhinal cortex. These regions contain cells involved in spatial navigation, and studying the statistical properties of their spiking activity is a key step to understanding biological navigation.