Adaptive Estimator Selection for Off-Policy Evaluation

Authors: Yi Su, Pavithra Srinath, Akshay Krishnamurthy

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also perform comprehensive experiments, demonstrating the empirical efficacy of our approach and comparing with related approaches.
Researcher Affiliation Collaboration 1Cornell University, Ithaca, NY 2Microsoft Research, New York, NY.
Pseudocode No The paper describes the SLOPE procedure in prose within the text, but it does not provide a clearly labeled 'Pseudocode' or 'Algorithm' block or figure.
Open Source Code Yes Code for this section is publicly available at https:// github.com/Vowpal Wabbit/slope-experiments.Code for this section is available at https://github. com/clvoloshin/OPE-tools.
Open Datasets Yes We use four RL environments: Mountain Car, Gridworld, Graph, and Graph-POMDP (abbreviated MC, GW, Graph, and PO-Graph). All four environments are from Voloshin et al. (2019)
Dataset Splits No The paper describes conducting experiments with a certain number of replicates and generating data for each condition (e.g., 'We perform 30 replicates of each condition'), but it does not specify explicit training, validation, or test dataset splits in percentages or absolute counts for a fixed dataset.
Hardware Specification No The paper does not specify any particular hardware components such as GPU or CPU models, memory, or cloud instance types used for running experiments.
Software Dependencies No The paper does not provide specific version numbers for any software libraries, frameworks, or dependencies used in the experiments (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes We consider 7 different choices of geometrically spaced bandwidths H := {2 i : i 2 [7]}. For SLOPE, we simplify the implementation by replacing the confidence function in (2), with twice the empirical standard deviation of the corresponding estimate. We also manually enforce monotonicity of this confidence function.