Unsupervised Learning by Program Synthesis

Authors: Kevin Ellis, Armando Solar-Lezama, Josh Tenenbaum

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We apply our techniques to both a visual learning domain and a language learning problem, showing that our algorithm can learn many visual concepts from only a few examples and that it can recover some English inflectional morphology.
Researcher Affiliation Academia Kevin Ellis Department of Brain and Cognitive Sciences Massachusetts Institute of Technology ellisk@mit.edu Armando Solar-Lezama MIT CSAIL Massachusetts Institute of Technology asolar@csail.mit.edu Joshua B. Tenenbaum Department of Brain and Cognitive Sciences Massachusetts Institute of Technology jbt@mit.edu
Pseudocode Yes Algorithm 1 SMT encoding of programs generated by production P of grammar G
Open Source Code No The paper does not provide any explicit statements or links indicating that the code for the described methodology is open-source or publicly available.
Open Datasets Yes We took five inflected forms of the top 5000 lexemes as measured by token frequency in the CELEX lexical inventory [15]." and "We take our visual concepts from the Synthetic Visual Reasoning Test (SVRT), a set of visual classification problems...
Dataset Splits No The paper states "We split this in half to give 2500 lexemes for training and testing" but does not explicitly provide details about a validation dataset split or a specific cross-validation setup for the general model training.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., GPU/CPU models, memory, number of cores) used for running the experiments.
Software Dependencies No The paper mentions software tools like "Z3" and "Morfessor" but does not provide specific version numbers for these or any other software dependencies crucial for replication.
Experiment Setup Yes Then, the x and y coordinates of each shape are perturbed by additive noise drawn uniformly from δ to δ; in our experiments, we put δ = 3." and "In our experiments, we put ϵ = 0.1." and "Concretely, we sampled many subsets of the data, each with 4, 5, 6, or 7 lexemes (thus 20, 25, 30, or 35 words)".