Fourier Representations for Black-Box Optimization over Categorical Variables

Authors: Hamid Dadkhahi, Jesus Rios, Karthikeyan Shanmugam, Payel Das10156-10165

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

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments over synthetic benchmarks as well as real-world RNA sequence optimization and design problems demonstrate the representational power of the proposed methods, which achieve competitive or superior performance compared to state-of-the-art counterparts, while improving the computation cost and/or sample efficiency, substantially.
Researcher Affiliation Industry Hamid Dadkhahi1*, Jesus Rios2, Karthikeyan Shanmugam2, Payel Das2 1 Amazon 2 IBM Research AI dadkhami@amazon.com, jriosal@us.ibm.com, karthikeyan.shanmugan2@ibm.com, daspa@us.ibm.com
Pseudocode Yes Algorithm 1: SA for Categorical Variables with Surrogate Model
Open Source Code No The paper does not provide an explicit statement or a link to open-source code for the described methodology.
Open Datasets Yes Synthetic Benchmarks: We consider two synthetic problems: Latin square problem (Colbourn and Dinitz 2006), a commonly used combinatorial optimization benchmark, and the pest control problem considered in (Oh et al. 2019) (see Appendix for the latter results)... In our experiments, we focus on three puzzles from the Eterna-100 dataset (Anderson-Lee et al. 2016).
Dataset Splits No The paper does not provide specific details about training, validation, or test dataset splits (e.g., percentages or counts). It describes evaluation over time steps.
Hardware Specification Yes The experiments are run on machines with CPU cores from the Intel Xeon E5-2600 v3 family.
Software Dependencies No The paper mentions using the "RNAfold package (Lorenz et al. 2011)" but does not specify a version number for this or any other software dependency.
Experiment Setup Yes The maximum degree of interactions used in our surrogate models is set to two for all the problems; increasing the max order improved the results only marginally (see Appendix). The sparsity parameter λ in exponential weight updates is set to 1 in all the experiments following the same choice made in (Dadkhahi et al. 2020)... The decay parameter used in the annealing schedule of SA is set to ℓ= 3 in all the experiments. In addition, the number of SA iterations T is set to 3 n and 6 n for ECO and TCO, respectively... we set the exploration parameter c to 0.5 for ECO and 0.25 for TCO; and the number of MCTS playouts at each time step to 30 h, where h is the tree height (i.e. number of dots and bracket pairs).