Welcome! I am currently a second year PhD student at the Computer Sciences Department of UW-Madison, where I am fortunate enough to be working with Prof. Dimitris Papailiopoulos. My current research interests revolve around Large Language Models, and both theoretical and practical aspects of Machine Learning and Deep Learning.
Before coming to Madison, I earned my BSc and integrated MEng from the ECE department of the Technical University of Crete. During this time, I had the opportunity to work with Prof. Aggelos Bletsas, focusing on asynchronous inference algorithms for ambiently powered wireless sensor networks [Diploma Thesis].
Currently looking for an internship for the summer of 2024.
This work offers design and implementation of in-network inference, using message passing among ambiently powered wireless sensor network (WSN) terminals. The stochastic nature of ambient energy harvesting dictates intermittent operation of each WSN terminal and as such, the message passing inference algorithms should be robust to asynchronous operation. It is shown, perhaps for the first time in the literature (to the best of our knowledge), a proof of concept, where a WSN harvests energy from the environment and processes itself the collected information in a distributed manner, by converting the (network) inference task to a probabilistic, in-network message passing problem, often at the expense of increased total delay. Examples from Gaussian belief propagation and average consensus (AC) are provided, along with the derivation of a statistical convergence metric for the latter case. A k-means method is offered that maps the elements of the calculated vector to the different WSN terminals and overall execution delay (in number of iterations) is quantified. Interestingly, it is shown that there are divergent instances of the in-network message passing algorithms that become convergent, under asynchronous operation. Ambient solar energy harvesting availability is also studied, controlling the probability of successful (or not) message passing. Hopefully, this work will spark further interest for asynchronous message passing algorithms and technologies that enable in-network inference, toward ambiently powered, batteryless Internet of Things-That-Think.
Is it possible to build ultra-low power wireless sensor networks (WSN) that exploit the inherent parallel and distributed nature of powerful message passing/inference algorithms, embrace ultra-low power communication principles and make autonomous, in-network decisions, solely powered by the environment? While edge and cloud computing emerge, this work points towards the opposite direction, inspired by the fact that ambient energy, either from radio frequency (RF), sun, motion, temperature or even living organisms, has fixed (on average) density per surface (or volume). It is shown, perhaps for the first time in the literature (to the best of our knowledge), a proof of concept, where a WSN harvests energy from the environment and processes itself the collected information in a distributed manner, by converting the (network) inference task to a probabilistic, message passing problem. Examples from Gaussian Belief Propagation and Average Consensus are offered; ambient energy harvesting and availability are quantified, controling the probability of successful (or not) message passing. Such interrupted communication requires distributed algorithms robust to asynchrony, at the expense of increased overall delay. Simulation and experimental validation are offered in a WSN testbed with solar energy harvesting. Future work will focus on overall delay minimization.