Learning Procedural Abstractions and Evaluating Discrete Latent Temporal Structure

Authors: Karan Goel, Emma Brunskill

ICLR 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 EXPERIMENTS WITH EVALUATION CRITERIA, 6 EXPERIMENTS WITH PRISM, Table 1: Comparison with baselines on BEES and the JIGSAWS surgical dataset., Table 2: Comparison with Sener & Yao (2018) on the INRIA dataset with the Munkres score.
Researcher Affiliation Academia Karan Goel Department of Computer Science Stanford University kgoel@cs.stanford.edu, Emma Brunskill Department of Computer Science Stanford University ebrun@cs.stanford.edu
Pseudocode Yes Algorithm 1 Calculation of RSS
Open Source Code Yes The results presented for the evaluation criteria developed in this paper can be reproduced using code available at https://github.com/Stanford AI4HI/ICLR2019_evaluating_ discrete_temporal_structure. and The results presented for PRISM can be reproduced using code available at https://github.com/Stanford AI4HI/ICLR2019_prism, which contains a Python implementation of PRISM.
Open Datasets Yes We use 2 common benchmark datasets (Fox et al., 2008b; 2009; 2014; Zhou et al., 2008; 2013). BEES consists of 6 time-series..., JIGSAWS dataset (Gao et al., 2014)..., Breakfast actions (Kuehne et al., 2014)..., INRIA instructional videos (Alayrac et al., 2016)...
Dataset Splits No No specific numerical train/validation/test splits (percentages or counts) or detailed splitting methodologies were explicitly stated in the paper. While datasets are mentioned, their partitioning for training, validation, and testing is not specified in a reproducible manner.
Hardware Specification No No specific details regarding the hardware specifications (e.g., CPU, GPU models, memory, or cloud instance types) used for running experiments were provided.
Software Dependencies No The paper mentions 'Python implementation of PRISM' but does not specify version numbers for Python or any other software dependencies, libraries, or frameworks used.
Experiment Setup Yes We use the hyperparameter settings given in Table 10 for PRISM. The Bayesian HMM has a single hyperparameter α which represents the hyperparameter for the Dirichlet prior over the transition matrix. and Table 10: Hyperparameter settings used for PRISM experiments.