Expectation-Maximization for Learning Determinantal Point Processes

Authors: Jennifer A Gillenwater, Alex Kulesza, Emily B. Fox, Ben Taskar

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

Reproducibility Variable Result LLM Response
Research Type Experimental We test our method on a real-world product recommendation task, and achieve relative gains of up to 16.5% in test log-likelihood compared to the naive approach of maximizing likelihood by projected gradient ascent on the entries of the kernel matrix.
Researcher Affiliation Academia Jennifer Gillenwater Computer and Information Science University of Pennsylvania jengi@cis.upenn.edu Alex Kulesza Computer Science and Engineering University of Michigan kulesza@umich.edu Emily Fox Statistics University of Washington ebfox@stat.washington.edu Ben Taskar Computer Science and Engineering University of Washington taskar@cs.washington.edu
Pseudocode Yes Algorithm 1 K-Ascent (KA) and Algorithm 2 Expectation-Maximization (EM) are presented.
Open Source Code Yes Code and data for all experiments can be downloaded from https://code.google.com/p/em-for-dpps
Open Datasets No To test our DPP learning algorithms, we collected a dataset consisting of 29,632 baby registries from Amazon.com, filtering out those listing fewer than 5 or more than 100 products.
Dataset Splits No The paper states: "We used 70% of the data for training and 30% for testing." It does not specify a validation split.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers.
Experiment Setup Yes The paper describes experimental setup details such as two initialization types (Wishart distribution and moments-matching), a 70% training and 30% testing data split, and data filtering criteria (e.g., 'filtering out those listing fewer than 5 or more than 100 products', 'filtered down to its top 100 most frequent items').