Danielle S. Bassett, University of Pennsylvania
How Humans Build Models of the World
Abstract: Human learners acquire not only disconnected bits of information, but complex interconnected networks of relational knowledge. The capacity for such learning naturally depends on the architecture of the knowledge network itself. I will describe recent work assessing network constraints on the learnability of relational knowledge, and a free energy model that offers an explanation for such constraints. I will then broaden the discussion to the generic manner in which humans communicate using systems of interconnected stimuli or concepts, from language and music, to literature and science. I will describe an analytical framework to study the information generated by a system as perceived by a biased human observer, and provide experimental evidence that this perceived information depends critically on a system’s network topology. Applying the framework to several real networks, we find that they communicate a large amount of information (having high entropy) and do so efficiently (maintaining low divergence from human expectations). Moreover, we also find that such efficient communication arises in networks that are simultaneously heterogeneous, with high-degree hubs, and clustered, with tightly-connected modules — the two defining features of hierarchical organization. Together, these results suggest that many real networks are constrained by the pressures of information transmission to biased human observers, and that these pressures select for specific structural features.
Bio: Prof. Bassett is the J. Peter Skirkanich Professor at the University of Pennsylvania, with appointments in the Departments of Bioengineering, Electrical & Systems Engineering, Physics & Astronomy, Neurology, and Psychiatry. Bassett is also an external professor of the Santa Fe Institute. Bassett is most well-known for blending neural and systems engineering to identify fundamental mechanisms of cognition and disease in human brain networks. Bassett is currently writing a book for MIT Press entitled Curious Minds, with co-author Perry Zurn Professor of Philosophy at American University. Bassett received a B.S. in physics from Penn State University and a Ph.D. in physics from the University of Cambridge, UK as a Churchill Scholar, and as an NIH Health Sciences Scholar. Following a postdoctoral position at UC Santa Barbara, Bassett was a Junior Research Fellow at the Sage Center for the Study of the Mind. Bassett has received multiple prestigious awards, including American Psychological Association’s ‘Rising Star’ (2012), Alfred P Sloan Research Fellow (2014), MacArthur Fellow Genius Grant (2014), Early Academic Achievement Award from the IEEE Engineering in Medicine and Biology Society (2015), Harvard Higher Education Leader (2015), Office of Naval Research Young Investigator (2015), National Science Foundation CAREER (2016), Popular Science Brilliant 10 (2016), Lagrange Prize in Complex Systems Science (2017), Erdos-Renyi Prize in Network Science (2018), OHBM Young Investigator Award (2020), AIMBE College of Fellows (2020).
Bassett is the author of more than 300 peer-reviewed publications, which have garnered over 26,000 citations, as well as numerous book chapters and teaching materials. Bassett is the founding director of the Penn Network Visualization Program, a combined undergraduate art internship and K-12 outreach program bridging network science and the visual arts. Bassett’s work has been supported by the National Science Foundation, the National Institutes of Health, the Army Research Office, the Army Research Laboratory, the Office of Naval Research, the Department of Defense, the Alfred P Sloan Foundation, the John D and Catherine T MacArthur Foundation, the Paul Allen Foundation, the ISI Foundation, and the Center for Curiosity.
Philippe Jacquet, Inria
Information Theory and Algorithms: The Frontiers of Artificial Intelligence. Toward a Theory of Learnability.
Abstract: We will investigate the connection between Information Theory (IT) and Artificial Intelligence (AI). The boundary between AI and IT is fuzzy. The main difference is that Information theory has never been a threat to human intelligence as some believe that AI can be. Stephen Hawking were among those concerned that AI might be able to supersede mankind intelligence. Information theory can answer, at least partially, to this emotional question. To make it short an isolated machine cannot evolve by itself to become more complex. To do so it must borrow entropy from the external world. We will develop an analogy with life evolution.
We will also investigate a theory of learnability: some problems are easier to learn than some other. For example, it takes a few iterations to a child to get the ability to recognize a cat in a picture, while some people will never be able to learn how to play music. In some case training may bring a neural network into a “swamp area” from where it may take an exponential time to escape. It is possible to mathematically characterize such swamp areas by the fact that the neural network has zero mean weights. Unfortunately many simple algorithmic problems lead to zero mean weight neural networks. It would imply that the control by AI of those system driven by basic algorithms would be difficult, and networks are among them. In other words there are areas where pure algorithms can still outperform AI and Machine Learning. We will finally quickly review some of the areas of resilience where the algorithms are still strongly needed before AI closes all the gaps.
Bio: Philippe Jacquet graduated from Ecole Polytechnique, Paris, France in 1981, and from Ecole des Mines in 1984. He received his PhD degree from Paris Sud University in 1989. Since 1998, he has been a research director in Inria, a major public research lab in Computer Science in France. He has been a major contributor to the Internet OLSR protocol for mobile networks. His research interests involve information theory, probability theory, artificial intelligence, protocol design, performance evaluation and optimization, and the analysis of algorithms. Since 2012 he is with Alcatel-Lucent Bell Labs as head of the department of mathematics of dynamic networks and information. Since 2019 he is with Inria as research director. He is recipient of the award Science et Defense 2004 from french government. He is IEEE Fellow.
Julia A. Schnabel, King’s College London
Abstract: Medical imaging spans the entire process from acquisition, reconstruction, and quality control to image segmentation, classification, and interpretation. Recent years have increasingly seen the use of machine learning and deep learning architectures along the entire imaging pipeline, providing innovative end-to-end learning solutions that can operate directly on the imaging sensor during image acquisition, for online interpretation by the clinician. In this talk I will focus on some recently developed “smart” medical imaging approaches applied to imaging problems in three major healthcare challenges: cancer, cardiovascular disease, and premature birth. I will specifically focus on physically and biologically realistic data augmentation, as well as real-time applications of our methods during scan-time, showing promise in image interpretation tasks that are typically only performed further down-stream, but that can equally contribute to achieving better image quality and more robust extraction of clinically relevant information.
Bio: Julia Schnabel graduated with an MSc in Computer Science at Technical University of Berlin (1993) and a PhD in Computer Science at University College London (1998), and subsequently held post-doctoral positions at University College London, King’s College London and University Medical Center Utrecht, before becoming first Associate Professor (2007) and then Full Professor (2014) of Engineering Science at the University of Oxford. In 2015 she joined King’s College London as Chair in Computational Imaging. Julia’s research focusses on machine/deep learning, complex motion modelling, as well as multi-modality and quantitative imaging for a range of medical imaging applications. She is serving on the Editorial Board of Medical Image Analysis, is Associate Editor for IEEE Transactions on Medical Imaging and IEEE Transactions on Biomedical Engineering, and has recently founded the new free open-access Journal of Machine Learning for Biomedical Imaging (melba-journal.org). She has been Program Chair of the MICCAI 2018 conference, is General Chair of IPMI 2021, and will be General Chair of MICCAI 2024, to be held for the first time in Africa. She is elected member of the IEEE EMBS Administrative Committee and the MICCAI Society Board of Directors, and an elected Fellow of the MICCAI Society (2018), ELLIS (2019), and IEEE (2021).