Random Tessellation Forests
Authors: Shufei Ge, Shijia Wang, Yee Whye Teh, Liangliang Wang, Lloyd Elliott
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present a simulation study, and analyse gene expression data of brain tissue, showing improved accuracies over other methods. In Section 3.1, we explore a simulation study that shows differences among u RTP and MRTP, and some standard machine learning methods. In Section 3.2, we examine predictions of a variety of RTF models for gene expression data. |
| Researcher Affiliation | Academia | 1Department of Statistics and Actuarial Science, Simon Fraser University, Canada 2School of Statistics and Data Science, LPMC & KLMDASR, Nankai University, China 3Department of Statistics, University of Oxford, UK |
| Pseudocode | Yes | Algorithm 1 Generative Process for RTPs and Algorithm 2 SMC for inferring RTP posteriors |
| Open Source Code | Yes | An implementation of our methods (released under the open source BSD 2-clause license) and a software manual are provided in the Supplementary Material. |
| Open Datasets | Yes | These datasets were acquired from NCBI s Gene Expression Omnibus1 and were are released under the Open Data Commons Open Database License. We provide test/train splits of the PCA preprocessed datasets in the Supplementary Material. 1Downloaded from https://www.ncbi.nlm.nih.gov/geo/ in Spring 2019. |
| Dataset Splits | No | In all of our experiments, for each train/test split, we allocate 60% of the data items at random to the training set. |
| Hardware Specification | No | We would also like to thank Fred Popowich and Martin Siegert for help with computational resources at Simon Fraser University. |
| Software Dependencies | No | We set the number of trees in all of the random forests to 100, which is the default in R s random Forest package [23]. |
| Experiment Setup | Yes | For all our experiments, we set the likelihood hyperparameters for the RTPs and RTFs to the empirical estimates αk to nk/1000. In all of our experiments, for each train/test split, we allocate 60% of the data items at random to the training set. We set the number of trees in all of the random forests to 100... For the all RTFs, we set the budget τ = . |