I'm not an expert in the field of systems neuroscience and I'm having a little trouble in formulating a concrete set of steps to answer a question I had this morning.
I'm interested in building an unsupervised classification algorithm for dynamical data based on the hypothesis that particular neurons learn to respond to particular characteristics in the input. Where would I begin investigating this issue?
I think if your goal is to build a classification algorithm, you don't need to build a detailed neuronal model, probably an artificial neural network suffices. You could have a look at so-called reservoir computing methods, for example. However, which method would be best depends on the particular problem you want to solve
My suggestion: contact Dr. Mauro Coppelli ([email protected]). He is a specialist who perhaps could help you.
There are three easy steps:
1. Find a good supervisor,
2. read all relevant literature,
3. Work.
Unsupervised learning rules in which neurons develop input specificity have been explored a lot in visual neuroscience. Here are two seminal works which present two unsupervised learning rules. When natural images are presented to these networks, neurons form 'receptive fields', and become selective to e.g. specific oriented bands in the visual input.
Olshausen, Bruno A. "Emergence of simple-cell receptive field properties by learning a sparse code for natural images." Nature 381.6583 (1996): 607-609.
Law, C. Charles, and Leon N. Cooper. "Formation of receptive fields in realistic visual environments according to the Bienenstock, Cooper, and Munro (BCM) theory." Proceedings of the National Academy of Sciences 91.16 (1994): 7797-7801.
- Badges
- Science topic
- Similar topics
- Biological Science
- Neuroscience