Automatic Construction and Natural-Language Description of Nonparametric Regression Models

Authors: James Lloyd, David Duvenaud, Roger Grosse, Joshua Tenenbaum, Zoubin Ghahramani

AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We compare ABCD against existing model construction techniques in terms of predictive performance at extrapolation, and we find stateof-the-art performance on 13 time series. We evaluate the performance of the algorithms listed below on 13 real time-series from various domains from the time series data library
Researcher Affiliation Academia James Robert Lloyd Department of Engineering University of Cambridge; David Duvenaud Department of Engineering University of Cambridge; Roger Grosse Brain and Cognitive Sciences Massachusetts Institute of Technology; Joshua B. Tenenbaum Brain and Cognitive Sciences Massachusetts Institute of Technology; Zoubin Ghahramani Department of Engineering University of Cambridge
Pseudocode No The paper does not contain any sections or figures explicitly labeled as 'Pseudocode' or 'Algorithm'.
Open Source Code Yes Source Code Source code to perform all experiments is available on github3. 3http://www.github.com/jamesrobertlloyd/gpss-research.
Open Datasets Yes We evaluate the performance of the algorithms listed below on 13 real time-series from various domains from the time series data library (Hyndman, Accessed summer 2013); plots of the data can be found at the beginning of the reports in the supplementary material.
Dataset Splits Yes As a heuristic, we order components by always adding next the component which most reduces the 10-fold cross-validated mean absolute error.
Hardware Specification No No specific hardware details (e.g., CPU, GPU models, or memory specifications) used for running experiments are mentioned in the paper.
Software Dependencies No The paper mentions using the 'GPML toolbox' for GP parameter optimisation but does not provide a version number for it or any other software dependencies.
Experiment Setup Yes After each model is proposed its kernel parameters are optimised by conjugate gradient descent. We evaluate each optimized model, M, using the Bayesian Information Criterion (BIC) (Schwarz, 1978): BIC(M) = 2 log p(D | M) + |M| log n. We use the default mean absolute error criterion when using Eureqa.