tlab and EXITO scholar Nancy reports from her 2022 COSYNE conference attendance.
This spring break, I attended the Computational and Systems Neuroscience conference (Cosyne) 2022 and workshops in Portugal – made possible by teuscher.:Lab and a grant from Cosyne. As an undergraduate, Cosyne set me and a group of other undergraduates up with two post-doc researchers in the field, Shashank Pisupati, and Ugurcan Mugan. They showed us the ropes of attending conference poster sessions, gave advice on graduate school and research, and generally were available to answer my questions and invite me into the community. This was my first in-person conference, while many people I talked to had been attending Cosyne for years. I was told that most researchers involved in computational neuroscience attended this conference.
Before the conference, I’d been following the work of G.R. Yang at The Center for Brains, Minds & Machines (CBMM) at MIT. I didn’t know he would be there, and then suddenly, he was sitting next to me at one of the speaking sessions. I was able to strike up a conversation and talk about the lab’s research with one of his graduate students. I also connected with other PIs, graduates, and post-docs about their work. Additionally, Cosyne arranged a meeting for our undergraduate group with some people from Google’s DeepMind, to get an industry perspective on research.
By attending Cosyne, I made valuable contacts, increased my knowledge of current research and techniques in neuroscience, gained more information on making decisions about graduate school, developed my own ideas on implementing computational models, and experienced Portugal.
Apply now for the 2022 Summer Graduate Research and Mentoring Program at https://www.teuscher-lab.com/alife14_biomolecules_workshop/grmp
Proctor, P.; Teuscher, C.; Hecht, A.; Osiński, M. Proximal Policy Optimization for Radiation Source Search. Journal of Nuclear Engineering, 2:368-397, 2021. https://doi.org/10.3390/jne2040029
Rapid search and localization for nuclear sources can be an important aspect in preventing human harm from illicit material in dirty bombs or from contamination. In the case of a single mobile radiation detector, there are numerous challenges to overcome such as weak source intensity, multiple sources, background radiation, and the presence of obstructions, i.e., a non-convex environment. In this work, we investigate the sequential decision making capability of deep reinforcement learning in the nuclear source search context. A novel neural network architecture (RAD-A2C) based on the advantage actor critic (A2C) framework and a particle filter gated recurrent unit for localization is proposed. Performance is studied in a randomized 20×20 m convex and non-convex simulation environment across a range of signal-to-noise ratio (SNR)s for a single detector and single source. RAD-A2C performance is compared to both an information-driven controller that uses a bootstrap particle filter and to a gradient search (GS) algorithm. We find that the RAD-A2C has comparable performance to the information-driven controller across SNR in a convex environment. The RAD-A2C far outperforms the GS algorithm in the non-convex environment with greater than 95% median completion rate for up to seven obstructions.