[Blog] W39. Spike finding W3.

Long term

Make me say: "Yeah, I think I have a satisfactory understanding of decision making in C-elegans". 

In a couple of years. Offer a framework for understanding and manipulating decision making on C elegans, and ensure that the approach is generalizable cross-species. Proposal.

In a year. Reproduce an in-vivo C elegans decision making study in computational model. Proposal.

Next week

Spike identification project (proposal, milestones)

• Two writeups shareable with Eli. [Why Vth-adjusted plots can be misleading, Doubts behind Vth's dependency on injected current]
• Analyze preliminary results on experiment 3.
• More spike finding setup. Repeat experiment 3 analysis on the non-linear I-V regions of experiment 2.
• Study AWA cell connectivity. What is special about the gap junctions and chemical synapses? Eli said there's something special but I didn't understand.

Reinforcement Learning on Dynamical Systems project

• Be able to explain the code behind this paper or any one paper that teammate recommends.

AMATH 534 project

• Find teammates, find 1 paper to reproduce, and have a sketch of the reproduction plan.

Past week

Spike identification project

• Completed experiment 2 writeup (changing steps for AWA) and discussed with Eli. Hysteresis of limit cycle was found, which I thought was exciting, but Eli is more interested in finding AWA spikes for now. Makes sense, since there is already existing literature, and the analysis is (probably) easier than explaining out hysteresis.
• Created a spreadsheet of future action items because now I have a bunch of them.
• Change in direction: From all-neuron bifurcation study to AWA spike. Read deeply into Kunert et al., 2017 to figure out the methods. But.., after talking with Eli, we decided reproducing it was not worth it because we want to focus on just 1 neuron, and do it at a very small bifurcation scale. New direction: Focus on reproducing AWA spike from Liu et al.
• Preliminary experiment 3 results: AWA spike finding. Made some plots using ranges that are close to Liu et al's setups. Haven't thought deeply about the results though.
• V-threshold dilemma. Dug into literature about the history behind Vthreshold (Vth)'s dependency on injected current, and why Vth-adjusted plots can be misleading. This is a model decision that I felt was strange, and I am doubting the usefulness of Vth-adjusted plots for this spike project. Will gather my thoughts and do two writeups on this next week.

Reinforcement Learning on Dynamical Systems project

• Read a review paper on DRL: A Brief Survey of Deep Reinforcement Learning.
• Found 1 paper that used DRL on partial differential equation. Skimmed, but didn't understand yet.

Education

• Gerstner lecture 7. Gerstner book chapters 5, 6.
• I don't think I'll be doing more Gerstner in the meantime. Not directly related to current projects, and I don't have the time.