Dear colleagues, We are looking for a highly motivated Ph.D. candidate with a background in machine learning and/or in electrical engineering for a project at the TU Delft at the crossroads of machine learning and hardware design. Building autonomous agents that can reliably compute and take decisions in noisy and uncontrolled environments is among the top research areas in today's artificial intelligence (AI). Yet, doing so within constrained power budgets for battery-operated edge devices is currently an open challenge. Indeed, while current cutting-edge deep-learning approaches can now reach acceptable performance in such environments, they are still subject to adversarial attacks and need a backend of GPU clusters, thereby requiring 4 to 6 orders of magnitude more power than is allowed for by the edge power budgets. On both ends of the spectrum of learning algorithms are the error backpropagation algorithm, i.e. the workhorse of modern deep learning, and local Hebbian learning rules, which are inspired by the brain's synaptic plasticity mechanisms. The former offers excellent performance but its energy/memory footprint is incompatible with low-power edge devices, while the latter allows for low-cost hardware implementations but can hardly be deployed beyond toy problems. In this PhD project, you will tackle this challenge by: * developing a theoretical framework that will allow exploring and generating, in a use-case-driven fashion, emerging learning algorithms that combine the strengths of local brain-inspired learning and backprop, * with a hardware/algorithm co-design approach, developing custom neuromorphic silicon prototypes (digital, then mixed-signal) for the proposed learning rules * investigating the deployment of these adaptive prototypes in resource-constrained use cases at the edge, such as brain implants for seizure detection. This project is a collaboration between Dr. Charlotte Frenkel (neuromorphic hardware, hardware/algorithm co-design, brain-inspired machine learning) and Dr. Justin Dauwels (Bayesian machine learning, computational neuroscience, biosignal processing). About the Department of Microelectronics at TU Delft: https://microelectronics.tudelft.nl/ About the CogSys research lab: https://ei.et.tudelft.nl/Research/theme.php?id=63 The expected starting date is 01/11/2023 and deadline for application is 01/09/2023. For more information and to apply, please visit this website <https://www.tudelft.nl/over-tu-delft/werken-bij-tu-delft/vacatures/details? jobId=13043&jobTitle=PhD%20Position%20in%20Bio-plausible%20Local%20Learning% 20Rules%20for%20Adaptive%20Neuromorphic%20Hardware> . Greetings, Justin. _______________________________________________________________ Justin Dauwels j.h.g.dauwels@tudelft.nl <mailto:j.h.g.dauwels@tudelft.nl> Associate Professor http://cas.tudelft.nl/ Fac. EEMCS Section Circuits and Systems Mekelweg 4 2628 CD Delft, The Netherlands _____________________________________________________________________