VaRT: Variational Regression Trees
Authors: Sebastian Salazar
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate the model s performance on 18 datasets and demonstrate its competitiveness with other state-of-the-art methods in regression tasks. |
| Researcher Affiliation | Academia | Sebastian Salazar Escobedo Department of Computer Science Columbia University New York, NY 10027 sebastian.salazar@cs.columbia.edu |
| Pseudocode | No | The paper describes processes and derivations but does not include a clearly labeled pseudocode or algorithm block. |
| Open Source Code | Yes | Code with the random seeds needed to replicate the results of this section are provided with the supplementary material. |
| Open Datasets | Yes | We conducted experiments on 18 distinct Datasets from the UCI Machine Learning Repository to benchmark our algorithm. |
| Dataset Splits | No | The RMSE values correspond to the average RMSE on the test sets of a random 90/10% train-test split of the data over ten runs. While early stopping by evaluating the RMSE on a training or validation set is mentioned as a possibility, the reported experiments explicitly state a train-test split without detailing a validation split. |
| Hardware Specification | Yes | We conducted all experiments on an ASUS Zephyrus G14 laptop with an RTX 2060 Max-Q GPU (6 GB VRAM), a 4900HS AMD CPU, 40 GB of RAM (although RAM usage was kept below 6 GB). |
| Software Dependencies | No | Va RT was trained using gradient descent paired with a Clipped Adam optimizer in Py Torch (Paszke et al. [2019], Bingham et al. [2018]). No specific version number for PyTorch or other software dependencies is provided. |
| Experiment Setup | Yes | A single Va RT tree for depths 3, 5, 7, and 10 were trained for each dataset... The regularization parameters for all runs was set to 10 3 and no hyperparameter tuning was performed to make a fair comparison on the off-the-shelf performance of each of the algorithms. |