On the Iteration Complexity of Oblivious First-Order Optimization Algorithms

Authors: Yossi Arjevani, Ohad Shamir

ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We consider a broad class of first-order optimization algorithms which are oblivious, in the sense that their step sizes are scheduled regardless of the function under consideration, except for limited side-information such as smoothness or strong convexity parameters. With the knowledge of these two parameters, we show that any such algorithm attains an iteration complexity lower bound of Ω( p L/ϵ) for L-smooth convex functions, and Ω( p L/µ ln(1/ϵ)) for Lsmooth µ-strongly convex functions.
Researcher Affiliation Academia Yossi Arjevani YOSSI.ARJEVANI@WEIZMANN.AC.IL Weizmann Institute of Science, Rehovot 7610001, Israel Ohad Shamir OHAD.SHAMIR@WEIZMANN.AC.IL Weizmann Institute of Science, Rehovot 7610001, Israel
Pseudocode Yes SCHEME 4.2 RESTARTING SCHEME PARAMETERS SMOOTHNESS PARAMETER L > 0 STRONG CONVEXITY PARAMETER µ > 0 CONVERGENCE PARAMETERS α > 0, C > 0 GIVEN A p-CLI OVER L-SMOOTH FUNCTIONS P WITH f(xk) f CL x0 x 2 FOR ANY INITIALIZATION VECTOR x0 ITERATE FOR t = 1, 2, . . . RESTART THE STEP SIZE SCHEDULE OF P INITIALIZE P AT x0 RUN P FOR αp 4CL/µ ITERATIONS SET x0 TO BE THE LAST ITERATE OF THIS EXECUTION END
Open Source Code No The paper is theoretical and does not mention releasing any source code for its described methodology.
Open Datasets No The paper is theoretical and does not involve empirical experiments using datasets, thus no information on dataset availability is provided.
Dataset Splits No The paper is theoretical and does not involve empirical experiments, so it does not discuss dataset splits for training, validation, or testing.
Hardware Specification No The paper is theoretical and does not describe any computational experiments, and therefore does not specify hardware used.
Software Dependencies No The paper is theoretical and does not describe any software implementation details or dependencies with version numbers.
Experiment Setup No The paper is theoretical and does not describe any empirical experimental setup, hyperparameters, or training configurations.