Artificial intelligence is transforming our world by solving tasks that were previously unthinkable, going from recognizing objects to translating text into other languages. The state-of-the-art algorithms underlying these novel capabilities are based on artificial neural networks or deep learning algorithms. Although extremely powerful, deep neural networks are often not efficient, requiring billions of computations for inference and training. In contrast, other machine learning algorithms, such as graphical model, support vector machines or spiking-neural-networks, can be more efficient but are often less powerful to model the complex relationships in real-life big data. This leads to the following question: Can we develop powerful, yet efficient, algorithms by combining deep learning algorithms with other machine learning methods?
In this PhD we will explore new ways to combine deep learning with other machine learning algorithms to achieve efficient machine learning on chip. We will examine how such hybrid models can enable efficient inference on hardware, e.g. by reducing the memory and computational requirements. Moreover, we will investigate how hybrid models can facilitate efficient learning on chip.
Our goal will be to make on-chip inference and training efficient, resulting in smart chips, which are flexible in their functionality and have the ability to fulfill a range of tasks on both mobile devices and in the cloud. To achieve these goals, we will leverage Imec's know-how in AI-accelerator design, novel memory and logic technology and algorithm development.
Required background: Computer science, AI/machine learning, hardware knowledge/interest
Type of work: 70% modeling, 20%hardware, 10% literature
Supervisor: Marian Verhelst
Daily advisor: Bram Verhoef
The reference code for this position is 2020-055. Mention this reference code on your application form.
Chinese nationals who wish to apply for the CSC scholarship, should use the following code when applying for this topic: CSC2020-22.