Pliable Rejection Sampling

Authors: Akram Erraqabi, Michal Valko, Alexandra Carpentier, Odalric Maillard

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

Reproducibility Variable Result LLM Response
Research Type Experimental We compared PRS to SRS and A sampling numerically. In particular, we evaluated the sampling rate, i.e., the proportion of samples that a method gives with respect to the number of evaluation of f. This is equal to the definition of acceptance rate for SRS and PRS. Figure 3 gives the acceptance rates of all these methods for a {2, 5, 10, 15, 20} averaged over 10 trials. Figure 3a corresponds to a budget of n = 10^5 requests to f and Figure 3b to a budget of n = 10^6 requests.
Researcher Affiliation Academia Akram Erraqabi AKRAM.ER-RAQABI@UMONTREAL.CA MILA, Universit e de Montr eal, Montr eal, QC H3C 3J7, Canada Michal Valko MICHAL.VALKO@INRIA.FR INRIA Lille Nord Europe, Seque L team, 40 avenue Halley 59650, Villeneuve d Ascq, France Alexandra Carpentier CARPENTIER@MATH.UNI-POTSDAM.DE Institut f ur Mathematik, Univert at Potsdam, Germany, Haus 9 Karl-Liebknecht-Strasse 24-25, D-14476 Potsdam Odalric-Ambrym Maillard ODALRIC.MAILLARD@INRIA.FR INRIA Saclay ˆIle-de-France, TAO team, 660 Claude Shannon, Universit e Paris Sud, 91405 Orsay, France
Pseudocode Yes Algorithm 1 Pliable rejection sampling (PRS)
Open Source Code No The paper thanks Chris Maddison for his code of A sampling, but it does not state that the authors are releasing their own code for PRS or the experiments described in the paper.
Open Datasets Yes We use two of the same settings in of Maddison et al. (2014). We first study the behavior of the acceptance rate with as a function of to the peakiness of f. In particular, we use the target density of Maddison et al. (2014), where a is the peakiness parameter. In order to illustrate how PRS behaves for inference tasks, we tested the methods on the clutter problem of Minka (2001) as did Maddison et al. (2014).
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning. The experiments involve sampling from a target density, not training a model with explicit dataset splits.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes All the experiments were run with δ = 0.01. HC was set through a cross-validation in order to provide a good proposal quality, i.e., how close is the proposal to the target distribution. Figure 3a corresponds to a budget of n = 10^5 requests to f and Figure 3b to a budget of n = 10^6 requests.