2022-11-30 Self Supervised Learning at the Edge: Addressing Label Scarcity on Online Data - Dario Dematties
Edge computing is a new computing paradigm that places powerful computers for analysis, inference, and learning close to physical data sources like sensors, breaking the historical "aggregate your data at a central location and compute" approach. When inference using a Machine Learning or Deep Learning model is demanded by a task at hand, the standard Artificial Intelligence at the Edge workflow consists of training a model offline using a well curated and labeled dataset, and then deploying the trained algorithm to make inference at the edge. Two challenges afflict this process. First, the data used to train the model is often different from the one collected by the edge device. This is due the fact that training models requires labeled datasets, and labeling new datasets is a time consuming and arduous process. Hence, often the research community at large uses a small number of labeled dataset from other similar data sources for training purposes. Second, machine learning inferences suffer from two broad aspects of Model drift - “concept drift” and “data drift”, which are further exacerbated in the AI@Edge.
Toward overcoming the above challenges, we have begun to explore Self Supervised Learning, with the goal to eventually learning at the edge. A viable option is to capture valuable information in the synaptic weights of an artificial neural network through a technique called Self Supervised Federated Learning. We are applying these techniques to two different datasets collected by Sage nodes - audio and images. In this talk I will show how our approach can autonomously extract relevant features from images of the sky - separating clouds and clustering them based on features, and separating silence and background noise from sounds of interesting events. This novel approach requires reduced human intervention, suggesting a new path for data characterization at the edge.
About the Speaker:
Dario Dematties is Postdoctoral Researcher at Northwestern-Argonne Institute of Science and Engineering.
Dario's research focuses on Self-supervised Learning (SSL) and Federated Learning (FL) for Edge Computing (EC) scenarios. He is applying SSL in the context of automatic atmospheric conditions characterization and in avian diversity monitoring. He has also collaborated with Uppsala University researchers in Sweden by applying Deep Learning to Nanopore Translocation Signal Processing.
Dario received his Electronic Engineering degree from Universidad Tecnológica Nacional (UTN) in Mendoza, Argentina in 2012 and his PhD degree at the Instituto de Ingeniería Biomédica, Facultad de Ingeniería, Universidad de Buenos Aires, Argentina in 2020. His PhD thesis focused on Biological Plausible modeling of early human language acquisition. Later at the National Scientific and Technical Research Council (CONICET) Argentina, he worked as a Postdoctoral Researcher on Bio-inspired Foveated Computer Vision using Deep Learning.
(Chicago Central time)
6:00 pm - Brief introductions
6:05 pm - Talks by Dario Dematties
6:50 pm - Q&A
7:00 pm - End
IMPORTANT: Complete both registrations:
Note: You must register with Zoom for the remote broadcast, registering in Meetup does not give you access to the Zoom meeting.