Lifelong Learning with Deep Neural Networks
Artificial agents interacting in highly dynamic environments are required to continually acquire and fine-tune their knowledge over time. In contrast to popular deep neural network models that typically rely on a large batch of annotated training samples, lifelong learning systems must account for situations in which the number of tasks is not known a priori and the data samples become incrementally available. Although significant advances have been made in domain-specific lifelong learning with artificial neural networks, extensive research efforts are required for the development of general-purpose artificial intelligence and autonomous agents which currently suffer from flexibility, robustness, and scalability issues. In this talk, I will introduce the main challenges linked to lifelong learning and discuss well-established research and recent methodological trends motivated by experimentally observed lifelong learning factors in biological systems.
German I. Parisi received his Bachelor's and Master's degree in Computer Science from the University of Milano-Bicocca, Italy. In 2017 he received his PhD in Computer Science from the University of Hamburg on the topic of multimodal action representations with deep recurrent neural networks. In 2015 he was a visiting researcher at the Cognitive Neuro-Robotics Lab at the Korea Advanced Institute of Science and Technology (KAIST), South Korea. Since 2017 he is a postdoctoral associate of the international research project Crossmodal Learning in the Knowledge Technology Institute at the University of Hamburg, Germany. His main research interests include neurocognitive systems, human-robot interaction and assistance, multimodal integration, and deep learning.