Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Experimental Design for Optimization of Orthogonal Projection Pursuit Models

Authors: Mojmir Mutny, Johannes Kirschner, Andreas Krause10235-10242

AAAI 2020 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate the algorithm numerically on synthetic as well as real-world optimization problems.
Researcher Affiliation Academia Mojm ır Mutn y ETH Zurich EMAIL Johannes Kirschner ETH Zurich EMAIL Andreas Krause ETH Zurich EMAIL
Pseudocode Yes Algorithm 1 Orthogonal PPR Bandit Algorithm Algorithm 2 Kernelized Thompson sampling Algorithm 3 Experimental Design for Hessian Estimation
Open Source Code No The paper does not provide an explicit statement or link to its open-source code for the described methodology.
Open Datasets Yes We validate our methods on standard benchmarks from the additive Bayesian optimization literature (Gardner et al. 2017). We first focus on an explanatory example in Figure 3a, where we optimize a two dimensional function. [...] We optimize a 5 dimensional function, which is a sum of polynomials of degree 4, where the polynomial kernel was used globally but due to sensitivity of misspecification (large Lipschitz constant), the squared exponential kernel was used along the coordinates. In the last benchmark problem (Figures 3b and 3c), which models the performance of a real-world electron laser machine
Dataset Splits No The paper references datasets and benchmarks but does not explicitly provide training/validation/test splits.
Hardware Specification No The paper does not explicitly describe the hardware used for its experiments.
Software Dependencies No The paper mentions 'pymanopt' but does not specify its version or other software dependencies with version numbers.
Experiment Setup Yes In practice, we specify the value of ϵ = 10 3 in the first phase of the algorithm, and we model TR separately as our analysis suggests larger (but not unreasonable) values for TR for short optimization horizons T.