PhD - Leuven | More than two weeks ago
Analog In Memory Computing has established itself as an energy efficient candidate for a wide range of Deep Learning Inference hardware. Current research shows the limitations of devices to be used for neural network inference and training due to several non-idealities. This PhD topic will focus on algorithms to enable analog computing based on novel memory technology for deep learning. The approach adopted for such a system would require co-optimization at every abstraction level of a computing system, starting from algorithm, architecture, circuits to device engineering.
Imec has a research program on developing new algorithms, systems and devices for deep learning. The PhD topic will enable the long-term vision of this research program to define and engineer algorithms and memory devices apt for Deep Learning. The student will be supported and supervised by a group of researchers working on enabling disruptive technology solution in the AI domain.
This PhD will build on imec's experience of combining system level constraints for technology or semiconductor innovation. The student will focus on the STCO techniques to identify, explore and solve the bottlenecks of engineering algorithms and devices for Analog-In-Memory-Compute systems. The research will involve fundamental breakthrough in algorithm level techniques coupled with novel circuit and device technology to find an energy efficient solution for running deep learning algorithms. This topic would be classic example of System-Technology-Co-Optimization philosophy pursued in imec.
Required background: fundamental knowledge in neural networks, device physics, basic computer architecture, basic circuit design skills would be preferred but not mandatory
Required background: Electrical Engineering, Nanotechnology, Computer Science
Type of work: 60% algorithm development and device modeling, 20% experimental, 10% literature
Supervisor: Ingrid De Wolf
Co-supervisor: Peter Vrancx
Daily advisor: Arindam Mallik, Jonas Doevenspeck
The reference code for this position is 2021-064. Mention this reference code on your application form.