Title: Stochastic neural architecture search and its application to medical images
Speaker: Dr. Leo Lim(Kakao Brain)
Date: 4pm, March. 7, 2019
Place: Math. Bldg. Rm. 404
Abstract:
Neural Architecture Search (NAS) algorithms lately attract attention from machine learning society due to their potential for Auto ML. The initial NAS algorithms employ reinforcement learning (RL) method which is powerful, but computationally expensive. In this lecture, it will be introduced that recent researches on NAS which are based on continuous relaxation e.g. Gumbel-softmax without using the RL method. We also verify the transferability of searched architecture by its application to medical images which consist of different modalities and organs.
|