Bandit Theory and Thompson Sampling-Guided Directed Evolution for Sequence Optimization

Authors: Hui Yuan, Chengzhuo Ni, Huazheng Wang, Xuezhou Zhang, Le Cong, Csaba Szepesvari, Mengdi Wang

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We finally harmonize our theoretical analysis with a set of simulation and real-world experiment. ... 6 Experiments ... 6.1 Simulation ... 6.2 Real-world experiment validation
Researcher Affiliation Academia 1,2,4,8Department of Electrical and Computer Engineering, Princeton University 3School of Electrical Engineering and Computer Science, Oregon State University 5Department of Pathology and Department of Genetics, Stanford University 6Department of Computing Science, University of Alberta
Pseudocode Yes Algorithm 1 Thompson Sampling-Guided Directed Evolution (TS-DE) ... Module 1 Crossover Selection(f, S) ... Module 2 Directed Mutation(f, S, µ)
Open Source Code No 3. If you ran experiments... (a) Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No]
Open Datasets No The paper describes simulation experiments starting with an 'initial population to be all zeros' and a 'real-world experiment validation' for 'optimizing a CRISPR design sequence', but it does not specify any publicly available dataset used for training with concrete access information.
Dataset Splits No The paper does not specify explicit training, validation, or test dataset splits. It describes simulation parameters like 'd = 10, T = 100 and µ = 0.8' and 'd = 40, M = 20 and µ = 0.1' but not data splits.
Hardware Specification No The paper does not provide specific hardware details. In the Ethics Statement, question 3(d) 'Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)?' is answered with '[N/A]'.
Software Dependencies No The paper does not list any specific software components with version numbers (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes We set the initial population to be all zeros, and set λ = 1, σ = 1. ... we plot the population-averaged Bayesian regret of TS-DE with various values of M, where d = 10, T = 100 and µ = 0.8. ... with d = 40, M = 20 and µ = 0.1.