AlgoExpert Competition
I'm making a JavaScript visualization of how plasticity alters the similarity between neural firing patterns. It's for a competition https://www.youtube.com/watch?v=KVyMIEwI7uw by Clem Mihailescu from AlgoExpert.io. His goal is to making a software engineering competition that would incentivize people to build cool things:-).
I thought about what to do and I have my simulation idea. When I saw this animation with d3js https://observablehq.com/@d3/force-directed-graph I thought yeah, that's exactly what I need. So I'll make artificial neuron nodes that receive as input data from real neurons and the connections between neurons will have tunable weights with different plasticity rules. A force based algorithm would then bring together neurons (nodes) that develop similar response properties. I need multiple plasticity rules to work together to get sensible results. I'll start with implementing in Python all the learning rules from Dayan and Abbott (a textbook on Theoretical Neuroscience).
I think it's my favorite passion, creating simulations of high-dimensional data. It's also what my job at Janelia is about, but there I code visualizations in PyQtGraph, which currently doesn't have great graph visualization opportunities. Once I'm done with a JavaScript implementation of the force based graph, I can consider contributing to PyQtGraph which is open source to make the algorithm available there:-)
You can track my progress on building the simulator with both data and simulated nodes and visualizations on my GitHub https://github.com/mariakesa/BonaFideSimulator
Hope you'll consider participating in the competition and building something cool:-)
I thought about what to do and I have my simulation idea. When I saw this animation with d3js https://observablehq.com/@d3/force-directed-graph I thought yeah, that's exactly what I need. So I'll make artificial neuron nodes that receive as input data from real neurons and the connections between neurons will have tunable weights with different plasticity rules. A force based algorithm would then bring together neurons (nodes) that develop similar response properties. I need multiple plasticity rules to work together to get sensible results. I'll start with implementing in Python all the learning rules from Dayan and Abbott (a textbook on Theoretical Neuroscience).
I think it's my favorite passion, creating simulations of high-dimensional data. It's also what my job at Janelia is about, but there I code visualizations in PyQtGraph, which currently doesn't have great graph visualization opportunities. Once I'm done with a JavaScript implementation of the force based graph, I can consider contributing to PyQtGraph which is open source to make the algorithm available there:-)
You can track my progress on building the simulator with both data and simulated nodes and visualizations on my GitHub https://github.com/mariakesa/BonaFideSimulator
Hope you'll consider participating in the competition and building something cool:-)
Kommentaarid
Postita kommentaar