Hey everyone! I’m really interested in using Brian for a ML task.
For example using it for classification.
Anyone have any ideas?
Or some done works on this?
Thanks a lot!
Spiking neural models are simulated using Brian2 and they can be used in a variety of analysis of neural data or to solve ML tasks.
There are papers that have used Spiking Neural Networks for classification like Iris flower classification or the standard MNIST. I am positively certain that Brian2 was used to simulate a Spiking Net model.
Here are some research papers:
Solving a classification task by spiking neurons with STDP and temporal coding
File Classification Based on Spiking Neural Networks
All the above papers solve typical ML tasks like classification using Spiking Neural networks, you can most certainly use Brian2 to reproduce the models presented in these papers.
Thank you so much!
Is there any sample code -Brian2 based- for these kind of ML tasks?
Another question that I have is about learning algorithm and I think that’s the tricky part!
Is there any built-in learning function in Brian2?
What are the different learning methods that I can implement them in my own?
I think there have been a few projects documented on this forum along these lines.
Most of the examples seem to use STDP, a way that neurons change the strength of their synpatic weights in biology.
However, please read through this discussion from @mstimberg on why STDP for learning static-representations may be missing the point of STDP in the brain
-
Brian 2 version source code for paper “Unsupervised Learning of Digit recognition using STDP”
- there’s a detailed Brian community discussion about updating this project here:
“Unsupervised learning of digit recognition using spike-timing-dependent plasiticity” - forum post
- there’s a detailed Brian community discussion about updating this project here:
some other external projects:
- “MNIST spiking neural network” - blogpost w/ code by Larionov
- “Three single-layer SNN learning algorithms with different homeostasis methods”
- “Investigation of Unsupervised Learning in Spiking Neural Network using STDP Rules”
- Machine Learning - Final Project - Schwartz & Liang