On Mixtures of Markov Chains

Authors: Rishi Gupta, Ravi Kumar, Sergei Vassilvitskii

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section we demonstrate that the algorithm is efficient, and performs well even when we use empirical observations. In addition, we also compare its performance against the most natural EM algorithm for the reconstruction problem. ... We use the last.fm 1K dataset3, which contains the list of songs listened by heavy users of Last.Fm.
Researcher Affiliation Collaboration Rishi Gupta Stanford University Stanford, CA 94305 rishig@cs.stanford.edu / Ravi Kumar Google Research Mountain View, CA 94043 ravi.k53@gmail.com / Sergei Vassilvitskii Google Research New York, NY 10011 sergeiv@google.com
Pseudocode No The paper describes the 'Reconstruction algorithm' in four high-level steps (i) Matrix decomposition, (ii) Co-kernel, (iii) Diagonalization, (iv) Two-trail matching, but does not present these steps as structured pseudocode or in a clearly labeled 'Algorithm' block.
Open Source Code No The paper does not contain any statements about releasing source code for the methodology or provide links to a code repository.
Open Datasets Yes We use the last.fm 1K dataset3, which contains the list of songs listened by heavy users of Last.Fm. ... 3http://mtg.upf.edu/static/datasets/last.fm/lastfm-dataset-1K.tar.gz
Dataset Splits No The paper states 'For each instance, we generate T samples of 3-trails' for synthetic data and 'We break each sequence into 3000 3-trails' for real data, but it does not specify any explicit training, validation, or test dataset splits, percentages, or cross-validation methodology for reproducibility of data partitioning.
Hardware Specification No The paper does not provide specific details on the hardware (e.g., CPU, GPU models, memory) used to run the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., library names with specific versions) used in the experiments.
Experiment Setup Yes For a specific n and L, we generate an instance (M, S) as follows. For each state i and Markov chain M ℓ, the set of transition probabilities leaving i is distributed as Dn. We draw each sℓfrom Dn as well, and then divide by L, so that the sum over all sℓ(i) is 1. ... For each instance, we generate T samples of 3-trails. ... In our implementation, we continue running EM until the log likelihood changes by less than 10 7 in each iteration; this corresponds to roughly 200-1000 iterations.