In this week’s biology seminar, we were very happy to have Dr. Jun-nosuke Teramae from Kyoto University to tell us about the mysterious stochastic behavior of neurons and synapses in the brain. Many biological experimental observations are reported. However, how this stochasticity is beneficial for computation and learning in the brain remains largely unknown. Dr. Teramae presented his work on developing an efficient learning algorithm inspired by this brain behavior. The algorithm is based on Gibbs-sampling which allow us to efficiently obtain high-dimensional sampling results. The algorithm shows similar stochastic behavior of the brain, which other machine leaning algorithms doesn’t show. In the end, Dr. Teramae briefly showed us the algorithm enables us to reproduce the recently discovered efficient power-law coding in the cortex. Even the seminar time is ended, the discussion continued for a long while. The seminar inspired many open questions on brain behavior and learning algorithms, both for the speaker and the audiences. We look forward to further development on the subjects. We thank Dr. Jun-nosuke Teramae and everyone joined the seminar.

Reported by Yingying Xu

関連イベント