Deep Learning has shown remarkable results in recent years in a wide range of computer vision tasks and novel network architectures have been proposed, However, with requirements for increased computational power and memory, it has become difficult to operate them using on-board LSI devices or edge devices. In this research, we are looking for a researcher who can use meta learning approaches to develop an automated model which can operate a deep neural network in an environment with limited computational resource such as an edge device, based on given limits. For example, we are looking for candidates who have brilliant vision and advanced skills to aim to build an automated deep neural network model architecture logic which based on limited expert data can complete a task. Additionally, using our facilities, software and data for research of the automated design of a neural network architecture, candidates will also explore the design of loss functions and optimization methods, and have the opportunity to work on their own related theme.
The aim of this research is to build the basics of a strong technology for Toshiba in the field of logistics, infrastructure, and self-driving cars.
Knowledge and Skills Required
Candidates are required to have a profound knowledge of at least one of the following fields: pattern recognition, machine learning, deep learning, reinforcement learning and robotics. They should have software implementation skills such as C, C++, C#, MATLAB(R) or python(TM).
 A. Yaguchi, et al. "Adam Induces Implicit Weight Sparsity in Recti?er Neural Networks," ICMLA 2018 (under review).
 K. Isogawa, et al. "Deep Shrinkage Convolutional Neural Network for Adaptive Noise Reduction," IEEE Signal Processing Letters, Vol.25, No.2, 2018.