Center for Advanced Studies (CAS) Research Focus Next Generation AI

Center for Advanced Studies (CAS) Research Focus Next Generation AI

By Center for Advanced Studies (CAS)Education
Download on the App Store

Center for Advanced Studies (CAS) Research Focus Next Generation AI episodes

  • On the Blending of Machine Learning and Economics
    Statistical decisions are often given meaning in the context of other decisions, particularly when there are scarce resources to be shared. Managing such sharing is one of the classical goals of microeconomics, and it is given new relevance in the modern setting of large, humanfocused datasets, and in data-analytic contexts such as classifiers and recommendation systems. Michael Jordan will discuss several recent projects that aim to explore the interface between machine learning and microeconomics, including the study of explorationexploitation trade-offs for bandit learning algorithms that compete over a scarce resource, leader/follower dynamics in strategic classification, and the robust learning of optimal auctions.
    1 hr 6 min
  • Content Algorithm Validation for Data Science
    Data Science (DS) algorithms interpret outcomes of empirical experiments with random influences. Often, such algorithms are cascaded to long processing pipelines especially in biomedical applications. The validation of such pipelines poses an open question since data compression of the input should preserve as much information as possible to distinguish between possible outputs. Starting with a minimum description length argument for model selection Joachim Buhmann motivates a localization criterion as a lower bound that achieves information theoretical optimality for sufficient statistics. Uncertainty in the input causes a rate distortion tradeoff in the output when the DS algorithm is adapted by learning. Joachim Buhmann presents design choices for algorithm selection and sketches a theory of validation.
    1 hr 29 min
  • The Impact of AlphaFold on Protein Research
    Karl-Peter Hopfner ist Professor für Biochemie am Genzentrum der LMU mit dem Fokus auf Strukturbiologie. | Julia Merlot ist Redakteurin im Wissenschaftsressort des SPIEGEL. | Alexander Pritzel ist wissenschaftlicher Mitarbeiter bei DeepMind und Mitentwickler des AlphaFold Programms
    1 hr 33 min
  • On the Blending of Machine Learning and Economics
    Statistical decisions are often given meaning in the context of other decisions, particularly when there are scarce resources to be shared. Managing such sharing is one of the classical goals of microeconomics, and it is given new relevance in the modern setting of large, humanfocused datasets, and in data-analytic contexts such as classifiers and recommendation systems. Michael Jordan will discuss several recent projects that aim to explore the interface between machine learning and microeconomics, including the study of explorationexploitation trade-offs for bandit learning algorithms that compete over a scarce resource, leader/follower dynamics in strategic classification, and the robust learning of optimal auctions.
    1 hr 6 min

About Center for Advanced Studies (CAS) Research Focus Next Generation AI

From the publisher's feed

We currently witness the impressive success of artificial intelligence (AI) in real-world applications, ranging from autonomous driving over speech recognition to the health care sector. At the same time, modern, typically data-driven AI methods have a similarly strong impact on science such as astronomy, physics, medicine – as well as humanities or social sciences, often replacing classical methods in the state of the art. In fact, at present, basically any research area is already impacted or starting to get involved in research questions in the realm of AI. However, despite this outstanding success, most of the research on AI is empirically driven and not only is a comprehensive theoretical foundation -- in particular, in the sense of explanations of decisions -- missing, but even the limitations of these methods are far from being well understood. It is also far from clear how AI-based methods can be optimally combined with classical methods based on physical models as domain knowledge.