Hello,
I will be presenting our new work, resource constrained neural network architecture search, on this Wednesday.
Title: resource
constrained neural network architecture search (will be appear on arXiv soon)
Authors: Yunyang
Xiong, Ronak Mehta and Vikas Singh
Presenter: Yunyang Xiong
Time: 4 pm, Wednesday, 12/5, 2018
Location: CS 3310
The design of neural network architectures is frequently either based on human expertise using trial/error and empirical
feedback or tackled via large scale reinforcement learning strategies. Most methods
in use require exorbitant computational resources.
The above challenges are exacerbated because the search procedure must now balance accuracy with certain budget
constraints on resources.
Our goal in this work is to formulate resource constrained architecture search as a set function optimization,
and design a heuristic based on ideas known
to be effective for constrained submodular optimization.
Best regards,
Yunyang