Ishai Ben-David

Ishai Ben-David is a researcher and educator noted for contributions to theoretical computer science and machine learning. He has held academic appointments and collaborative roles at leading research institutions and has published on topics spanning learning theory, algorithmic stability, and robustness. His work intersects with developments in complexity theory, statistical learning, and optimization.

Early life and education

Ben-David received formative training in mathematics and computer science, studying at institutions with strong traditions in theoretical research. During his undergraduate and graduate years he was exposed to curricula and mentors associated with Hebrew University of Jerusalem, Tel Aviv University, and research groups connected to international centers such as Massachusetts Institute of Technology and Stanford University. His doctoral studies involved interactions with faculty and researchers who have ties to classical topics represented by Vladimir Vapnik, Leslie Valiant, and the community around the Conference on Learning Theory. Early influences included exposure to seminars and workshops hosted by organizations like the Association for Computing Machinery, the IEEE, and regional research consortia.

Academic career

Ben-David's academic trajectory includes faculty and visiting positions at universities and research labs known for work in theoretical foundations. He has taught courses related to computational learning theory at departments affiliated with Carnegie Mellon University, University of California, Berkeley, and institutes that collaborate with centers such as Microsoft Research, Google Research, and national funding bodies like the European Research Council. His collaborations have spanned coauthors from groups at Princeton University, University of Cambridge, École Normale Supérieure, and interdepartmental centers linking Harvard University with engineering schools. He has served on program committees for conferences including the NeurIPS, COLT, and ICML, and held editorial roles connected to journals overseen by publishers like Springer and ACM.

Research contributions

Ben-David has contributed to a range of theoretical problems that connect algorithmic learning with sample complexity, domain adaptation, and adversarial robustness. His analyses build on foundational results from Vapnik–Chervonenkis theory and draw on tools originally developed by researchers such as Donald Knuth and Alan Turing in algorithmic analysis contexts. Specific themes in his work include quantifying generalization gaps in settings related to transfer learning studied in venues like ALT and exploring stability notions related to regularization techniques common in support vector machine literature. He has produced theoretical characterizations of when learning algorithms succeed under distribution shift, relating to problems addressed by scholars at Caltech and Columbia University.

Ben-David's papers examine connections between empirical processes and optimization landscapes, integrating perspectives from researchers at MIT and ETH Zurich who study stochastic gradient methods and convexity. He has analyzed the role of complexity measures—such as Rademacher complexity and covering numbers—in bounding learning risk, building on prior contributions from Peter Bartlett and Shai Shalev-Shwartz. Additionally, his work touches on algorithmic fairness and robustness, intersecting with discussions advanced at FAT* and policy-oriented forums hosted by Stanford Cyber Policy Center.

Awards and honors

Throughout his career, Ben-David has received recognition from academic societies and research programs. Awards and honors include distinctions from regional science foundations, invitations to deliver keynote and plenary talks at conferences like COLT and NeurIPS, and research fellowships associated with institutions such as the Sloan Foundation and national academies. He has been a recipient of competitive grants administered by agencies comparable to the National Science Foundation and has participated in collaborative projects funded by consortia including technology partnerships with DARPA-style programs and European initiatives supported by the Horizon 2020 framework.

Selected publications

- Ben-David, I., et al. Works addressing transfer learning, domain adaptation, and sample complexity published in proceedings of COLT, NeurIPS, and ICML. - Ben-David, I., research on stability and generalization appearing in journals and anthologies associated with Springer and IEEE venues. - Ben-David, I., collaborative pieces exploring theoretical bounds in adversarial settings presented at workshops co-located with ICLR and ALT. - Ben-David, I., surveys and tutorial-style writings for summer schools and lecture series organized by CWI, Simons Institute, and leading university programs.

Category:Theoretical computer scientists Category:Machine learning researchers