IDLab is a research group of Ghent University, as well as a core research group of imec. IDLab performs fundamental and applied research on data science and internet technology, and counts over 300 researchers. Our major research areas are machine learning and data mining; semantic intelligence; multimedia processing; distributed intelligence for IoT; cloud and big data infrastructures; wireless and fixed networking; electromagnetics, RF and high-speed circuits and systems.
IDLab is also part of imec, the world-leading research and innovation hub in nanoelectronics and digital technologies. The combination of our widely acclaimed leadership in microchip technology and profound software and ICT expertise is what makes us unique. By leveraging our world-class infrastructure and local and global ecosystem of partners across a multitude of industries, we create ground-breaking innovation in application domains such as healthcare, smart cities and mobility, logistics and manufacturing, and energy.
Various forms of neural networks (CNN, LSTM, spiking neural networks) have realized important breakthroughs in a myriad of application domains. The dominant model today is to train neural networks in the cloud on a large set of labelled data in a hardware-agnostic way. Afterwards, inference either always occurs in the cloud, or in some specific cases, the neural network is compressed, pruned, and quantized for mapping on lower footprint devices. This current model does not allow for efficient online learning on the device itself. There are however several dynamic environments where additional online learning and on-device learning capabilities are required to discover new categories and tasks to adapt a task to the specific environment.
We are looking for an excellent PhD student to work on on-device, online learning techniques. Fundamental research questions relate to i) detecting when a neural network should retrain; and ii) how to retrain such a network within a limited budget of power and/or labelled data.
Various options can be explored in the domains of spiking neural networks as well as in the domain of artificial neural networks. Spiking neural networks hold promise to combine hardware-efficiency with online learning capabilities, and prototype hardware platforms start to appear. In the domain of Artificial Neural Networks, there is increasing interest in techniques such as quantized training techniques, modular reconfigurable architectures and Bayesian neural networks.
We offer a fully funded PhD scholarship for a maximal period of 4 years (upon positive progress evaluation). The PhD research is fundamental and innovative, but with clear practical applications. You will join a young and enthusiastic team of researchers, post-docs and professors. The PhD position is immediately available.
Apply with motivation letter, scientific resume, abstract of your master thesis, diplomas and detailed academic results (courses and grades), relevant publications, and at least one reference contact. This information, as well as possible questions, must be sent to Prof. Pieter Simoens at firstname.lastname@example.org. After the first screening, suitable candidates will be invited for an interview (also possible via Skype) and may get a small assignment. Applications will be screened as soon as they are received. The position is open until the vacancy is filled.Continue reading
|Title||PhD in hardware-efficient techniques for on-chip online learning with neural networks|
|Job location||Kapeldreef 75, B-3001 Heverlee|
|Published||August 17, 2018|
|Application deadline||Unspecified deadline|
|Job type||PhD  |
|Fields||Artificial Neural Network,   Computer Communications (Networks),   Applied Mathematics,   Computational Sciences,   Computational Mathematics,   Machine Learning,   Electronics  |