A Spectral Approach to Item Response Theory

Authors: Duc Nguyen, Anderson Ye Zhang

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on synthetic and real-life datasets, ranging from small education testing datasets to large recommendation systems datasets show that our algorithm is scalable, accurate, and competitive with the most commonly used methods in the literature.
Researcher Affiliation Academia Md. Naimul Hoque, Department of Computer and Information Science, University of Pennsylvania, mdnguyen@seas.upenn.edu; Anderson Y. Zhang, Department of Statistics and Data Science, University of Pennsylvania, ayz@wharton.upenn.edu
Pseudocode Yes Algorithm 1 Spectral Estimator; Algorithm 2 Accelerated Spectral Estimator
Open Source Code Yes We include the python implementation of our spectral algorithm in the supplementary materials.
Open Datasets Yes We mention here a few notable datasets: RIIID [1] (m = 6k, n = 23k, education testing dataset), ML-20M [28] (m = 27k, n = 138k), Book-Genome [31] (m = 10k, n = 350k).
Dataset Splits No For tuning the prior distribution for MMLE, we select the prior distribution that admits the highest log-likelihood on a validation set. However, the paper does not specify the methodology or percentages for the data splits.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments, such as CPU or GPU models, or memory specifications.
Software Dependencies No The paper mentions 'python implementation' and references to external open source implementations for baselines, but does not specify its own software dependencies with version numbers.
Experiment Setup No The paper states that CMLE, JMLE, and the spectral method 'requires minimal model tuning' but does not provide specific hyperparameter values, optimizer settings, or detailed training configurations for reproduction.