EE Seminar: Inference Under Memory Constraints
הרישום לסמינר יבוצע באמצעות סריקת הברקוד למודל (יש להיכנס לפני כן למודל, לא באמצעות האפליקציה)- הרישום מסתיים ב- 15:10
Registration to the seminar will be done by scanning the barcode for the Moodle (Please enter ahead to the Moodle, NOT by application)- Registration ends at 15:10
Electrical Engineering Systems Seminar
Speaker: Tomer Berg
Ph.D. student under the supervision of Prof. Ofer Shayevitz and Prof. Or Ordentlich
Wednesday, 20th August 2025, at 15:00
Room 011, Kitot Building, Faculty of Engineering
Inference Under Memory Constraints
Abstract
The problem of statistical inference in its various forms has been the subject of decades-long extensive research. Most of the effort has been focused on characterizing the behavior as a function of the number of available samples, with far less attention given to the effect of memory limitations on performance. Recently, this latter topic has drawn much interest from the engineering and computer science literature.
In this talk, we will go over several canonical problems in the field of statistical inference under memory constraint, including hypothesis testing, parameter estimation, and distribution property testing / estimation. We will also identify recurrent themes, extract some fundamental building blocks for algorithmic construction, and overview useful techniques for lower bound derivations.
Specifically, in the first part of the talk, I will introduce the general inference problem and the notion of memory constrained algorithms (represented by finite-state machines), discuss several assumptions that can be made, and consider a brief overview of common methodologies and ideas used in deriving performance bounds in finite-memory inference.
In the second part of the talk, I will present our main contributions for several memory-limited inference problems, progressing from the classical and relatively simple problem of binary hypothesis testing, to the problem of continuous parameter estimation, to the more complex and contemporary problems of distribution property testing / estimation. I will contextualize our results within the memory-limited literature and highlight our contributions to the field. Finally, we discuss some conclusions and present interesting open problems for future research.
