Complete Bottom-Up Predicate Invention in Meta-Interpretive Learning

Authors: Céline Hocquette, Stephen H. Muggleton

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 6 Experiments 6.1 Research Hypotheses We experimentally test within this section whether the use of bottom-up iterations can improve learning performance. [...] Results. Accuracy results are presented in Figure 3a: a T-test suggests that the difference in accuracy is statistically significant (p < 0.05) for a training set up with up to 16 instances thus refuting Null Hypothesis 1. Sample complexity results are detailed in Figure 3c and show that performing bottom-up iterations can reduce the sample complexity.
Researcher Affiliation Academia Celine Hocquette, Stephen H. Muggleton Department of Computing, Imperial College London, London, UK {celine.hocquette16, s.muggleton}@imperial.ac.uk
Pseudocode Yes Algorithm 1 Bottom-Up Learner Input: second-order logic program B|E related to training examples E, number of iterations k, definitions of initial predicate H0 Output: logic program H
Open Source Code Yes The code for reproducing the experiments is available at https://github.com/celinehocquette/bottom_up.git
Open Datasets Yes We consider 94 real-world string transformation problems evaluated in [Cropper, 2019] and inspired from [Lin et al., 2014; Gulwani, 2011]. The dataset contains 10 positive examples for each problem.
Dataset Splits No The paper describes training and testing sets, and the split between them, but does not explicitly mention the use of a separate validation set or its split.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments.
Software Dependencies No The paper mentions the MIL system Metagol and a Prolog meta-interpreter but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes We perform between 1 and 3 bottom-up iterations. Initial facts are built from a sample of size 1 of the positive examples. We set the size of the top-down learner search space to n [2, 4] clauses. For k [0, 3], k bottom-up iterations are performed. [...] A timeout is set to 10 minutes.