Neuromatch 3 (NM3) happened a few weeks ago in virtual space, and it was great! There were hundreds of presentations over five days (October 26-30), many of them already archived on YouTube. This version of Neuromatch took the place of Society for Neuroscience (SfN), which was cancelled due to COVID. In this sense, Neuromatch is proving itself to be an improvement on the legacy conference. Between my two research groups (Representational Brains and Phenotypes and DevoWorm), we had five presentations submitted to NM3. Let's go through them one by one.
Contrast Between Biological and Artificial Neural Networks (BNNs vs ANNs). Download slides from Figshare. Video on YouTube. Click to enlarge.
This presentation was by Krishna Katyal and myself. Krishna is a regular contributor to the DevoWorm group. This presentation demonstrates several contrasts between Biological and Artificial Neural Networks, and how level of abstraction, network structure, and energetics all play a role in distinguishing the information-processing marvels known as biological brains.
Approaching Artificial Intelligence as an Embodied Developmental Process. Download slides from Figshare. Video on YouTube. Click to enlarge.
Thinking more along the lines of biological development, the next presentation features work done in the Representational Brains and Phenotypes Group. Taking a developmental approach to the classic Braitenberg Vehicle, we demonstrate how embodied developmental principles can be used to shape and guide networks for learning. Topics such as developmental contingency and the difference between morphogenesis and learning during the development of an artificial nervous system were also discussed.
The Psychophysics of Non-neuronal Cognition. Download slides from Figshare. Video on YouTube. Click to enlarge.
Are the advantages of cognitive information processing limited to organisms with a brain? This seems like a strange question, but can actually help us understand what a brain does and why it is important for coordinating the behavior of complex multicellular systems. In this presentation, which includes contributors to both the Representational Brains and Phenotypes and DevoWorm groups, we reconsider a model of Diatom movement as cognitive information processing. We also propose a series of potential scenarios in which this information processing occurs as well as Psychophysical-like measures to quantify these phenomena.
This abstract was submitted by Jesse Parent and Anson Lim, two regular contributors to the Orthogonal Research and Education Lab. Jesse is also a group leader and community manager in the lab. While they were not able to present during the scheduled time, they continue to work on this topic under the new and emerging Cognition Futures project.
The final talk was by Akshara Gopi (a regular contributor to the Orthogonal Research and Education Lab), along with myself and Ashwin Irungovel. This presentation was on convergence insufficiency in human vision, a topical specialty of Akshara and Ashwin. My contribution was to propose an agent-based model for this phenomenon.
Check out all of the great talks at NM3, including one by Rishabh Chakrabarty (regular research contributor to the Orthogonal Lab) and his co-author called "Seeing through the Mind’s Eye: reconstruction of the visual stimuli using 3D Generative-Adversarial Modeling". This intriguingly titled talk features research that combines Neuroimaging data with Deep Learning.
If you are interested in these topics and want to be involved in this research, join us at our weekly research meetings (Saturday Morning NeuroSim and DevoWorm group), or watch them on YouTube! I also invite you to join the Orthogonal Research and Education Lab Slack or OpenWorm Slack (DevoWorm) for continued discussion. Hope to see everyone at Neuromatch 4!
Your blog is a source of enlightenment. Thank you! Also check the louisiana gun laws for non residents
ReplyDelete