Just-In-Time Learning for Fast and Flexible Inference
Authors: S. M. Ali Eslami, Daniel Tarlow, Pushmeet Kohli, John Winn
NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results in Sec. 6 show that for general graphical models, our approach leads to significant improvements in inference speed (often several orders of magnitude) over importance sampling whilst maintaining overall accuracy, even boosting performance for models where hand designed EP message-passing operators are available. |
| Researcher Affiliation | Industry | S. M. Ali Eslami, Daniel Tarlow, Pushmeet Kohli and John Winn Microsoft Research {alie,dtarlow,pkohli,jwinn}@microsoft.com |
| Pseudocode | No | The paper does not contain explicit pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement or link to its own open-source code for the described methodology. It mentions using Infer.NET [12] and Stan [19], which are third-party tools. |
| Open Datasets | Yes | We obtain yield data for 10% of US counties for 2011 2013 from the USDA National Agricultural Statistics Service [1] and corresponding temperature data using [18]. |
| Dataset Splits | Yes | We split the observed message sets into training (70%) and hold out (30%), and train and evaluate the random forests using the two datasets. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. |
| Software Dependencies | Yes | The experiments were performed using the extensible factor API in Infer.NET [12]. [...] Infer.NET 2.5, 2012. Microsoft Research Cambridge. Website URL: http://research.microsoft.com/infernet. |
| Experiment Setup | Yes | Unless stated otherwise, we use default Infer.NET settings (e.g. for message schedules and other factor implementations). We set the number of trees in each forest to 64 and use quadratic regressors. |