Budgeted Optimization with Constrained Experiments

Authors: Javad Azimi, Xiaoli Fern, Alan Fern

JAIR 2016 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our proposed methods for both settings using synthetic and real functions. The experimental results demonstrate the efficacy of the proposed methods. [...] 6. Experimental Results
Researcher Affiliation Collaboration Javad Azimi EMAIL Microsoft, Sunnyvale, CA, USA; Xiaoli Z. Fern EMAIL School of EECS, Oregon State University; Alan Fern EMAIL School of EECS, Oregon State University
Pseudocode Yes Algorithm 1 The Greedy Non-Sequential Algorithm [...] Algorithm 2 Accelerated Greedy Algorithm
Open Source Code No No explicit statement or link for the open-sourcing of the code described in this paper was found.
Open Datasets Yes The first three functions: Cosines, Rosenbrock, and Discontinuous are benchmarks that have been widely used in previous studies on stochastic optimization (Anderson, Moore, & Cohn, 2000; Brunato, Battiti, & Pasupuleti, 2006; Azimi et al., 2010). [...] For the former we utilize data collected as part of a study on biosolar hydrogen production (Burrows, Wong, Fern, Chaplen, & Ely, 2009)
Dataset Splits No The paper describes using test functions and a set of initial random points for policy evaluation, but does not specify conventional train/test/validation dataset splits for models or evaluation.
Hardware Specification No The paper mentions 'un-optimized matlab implementation' and describes run times, but does not provide specific hardware details (e.g., CPU/GPU models, memory, or machine specifications) used for running experiments.
Software Dependencies No The paper mentions using 'Gaussian process' and an 'un-optimized matlab implementation' but does not provide specific version numbers for any software, libraries, or solvers used to replicate the experiment.
Experiment Setup Yes In this paper we set κ = 0.02 and signal variance σf = y2max [...] We evaluate our proposed approaches considering three different slope values; slope = 0.1, 0.15, 0.30. [...] we divide each input dimension into 100 equal-length subintervals. [...] Each run starts with n = 5 randomly selected initial points [...] we fixed the total budget to B = 15 and examine the effect of the cost-model slope parameter over values 0.1, 0.15 and 0.3. In later experiments, we will consider larger budgets.