HOME
RESEARCH
PUBLICATIONS
ABOUT & CV
|
Ioannis Kontoyiannis Churchill Professor of Mathematics of Information Statistical Laboratory, University of Cambridge Ó Email |
|
|
My research lies at the intersection of information theory, probability and statistics. I study how information can be quantified, compressed, efficiently transmitted, and extracted from data. Also, I work on how information-theoretic ideas can help understand the structure of randomness, communication algorithms, and AI systems. With collaborators, my work has included entropy estimation, the spectral theory of Markov processes, precise limits of data compression and coding, machine learning and optimization algorithms, and Bayesian theory and methods for learning from empirical data. A continuing interest is the connection between mathematical theory and statistical computation: Understanding both what is fundamentally possible and how it can be achieved algorithmically. Recent projects concern the sample complexity of data compression and channel coding, entropy inequalities in general abelian groups, connections with additive combinatorics and polarization, Bayesian causal discovery, and the convergence and efficiency of stochastic machine learning algorithms. These directions extend my longstanding interests in compression, probabilistic approximation and inference from dependent data. I am also a co-investigator in the EPSRC-funded INFORMED-AI Hub, which studies the information-theoretic foundations of distributed artificial intelligence. |
|
Research interests ● Theory and
algorithms for data compression ● Entropy and
information estimation ● Fundamental
limits of compression and communication ● Markov
processes ● Finite-sample bounds in information theory ● Information-theoretic
approaches to probabilistic results ● Entropy
inequalities and functional inequalities ● Additive
combinatorics ● Mathematical
foundations of AI ● Machine
learning and optimization ● Bayesian
theory, methodology and time series ● Independence
and conditional independence testing |
Applications ● Neuroscience ● Cybersecurity ● Finance ● Genomics ● Satellite
imaging ● Public
health Some current research themes ● Sample
complexity of source and channel coding ● Entropy
inequalities in probability and combinatorics ● Bayesian
inference and causal discovery ● Markov
processes, learning and optimization |