Lifelong Learning at the Edge: Algorithms and Foundations

Lifelong Learning (or Lifelong Machine Learning) is an advanced Machine Learning (ML) paradigm: an architecture for ML systems that learn continuously, accumulate knowledge learned in the past, and use that knowledge to help future learning and problem solving. This talk will explore three aspects of Lifelong Learning.

First, traditional ML is generally an energy-hungry process: the process of building and training of the model requires huge computational resources. It might be expected that Lifelong Learning might be limited to cloud and off-line computing only. But this talk will explain some ways that Lifelong Learning avoids that constraint – by introducing a signal propagation approach that combines learning into the inference process, enabling the system to add classes to its knowledge model at the same time it is being used to solve problems.

Second, most ML approaches are limited in their ability to utilize temporal patterns in its computation. But in Lifelong Learning, there are some new mechanisms that can make the AI temporally aware, mechanisms which can improve the system’s ability to work effectively with limited data as well as a significant savings in size and power usage.

Third, Lifelong Learning research continues to explore new models of computation – going beyond the traditional Turing computation model. Lifelong Learning researchers have been investigating “Super-Turing computation,” a model of computation that resembles biological learning.

Speaker(s): Hava Siegelmann

Room: Room 105, Bldg: Computer Science Building, 35 Olden St., Princeton University, Princeton, New Jersey, United States, 08544, Virtual: https://events.vtools.ieee.org/m/375905