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].

Active Search in Intensionally Specified Structured Spaces

Authors: Dino Oglic, Roman Garnett, Thomas Gaertner

AAAI 2017 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To study the empirical performance in silico, i.e., without conducting lab experiments, we design synthetic testbeds that share many characteristics with drug design (Section 4).
Researcher Affiliation Academia EMAIL Institut f ur Informatik III Universit at Bonn, Germany Roman Garnett EMAIL Dep. of Computer Science & Eng. Washington University in St. Louis, USA Thomas G artner EMAIL School of Computer Science The University of Nottingham, UK
Pseudocode Yes Algorithm 1 gives a pseudo-code description of our approach.
Open Source Code No The paper does not include any explicit statement or link indicating that the source code for the methodology is publicly available.
Open Datasets Yes We have simulated Algorithm 1 with the uniform proposal generator over the space of graphs with 7 and 10 nodes (Wormald 1987). For the space of cocktails, we have developed a frequency based sampler from a small set of cocktails collected from www.webtender.com.
Dataset Splits Yes To allow for models of varying complexity, we have estimated the conditional exponential family regularization parameter in each round using 5-fold stratified cross-validation.
Hardware Specification No The paper mentions 'University of Nottingham High Performance Computing Facility' but does not provide specific details such as GPU/CPU models or memory specifications used for experiments.
Software Dependencies No The paper mentions methods like 'k-NN active search' and 'decision trees' but does not provide specific software names with version numbers for reproducibility.
Experiment Setup Yes The Metropolis Hastings sampling was performed with a burn-in sample of 50 000 proposals and sampling was done for 50 rounds/batches. In each round we take 10 i.i.d. samples by running 10 Metropolis Hastings chains in parallel (note that samples from different rounds are dependent). To allow for models of varying complexity, we have estimated the conditional exponential family regularization parameter in each round using 5-fold stratified cross-validation.