NAOMI: Non-Autoregressive Multiresolution Sequence Imputation
Authors: Yukai Liu, Rose Yu, Stephan Zheng, Eric Zhan, Yisong Yue
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct exhaustive experiments on benchmark sequence datasets including traffic time series, billiards and basketball trajectories. Our method demonstrates 60% improvement in accuracy and generates realistic sequences given arbitrary missing patterns. |
| Researcher Affiliation | Collaboration | Yukai Liu Caltech Rose Yu Northeastern University Stephan Zheng Caltech, Salesforce Eric Zhan Caltech Yisong Yue Caltech |
| Pseudocode | Yes | Algorithm 1 Non-Aut Oregressive Multiresolution Imputation |
| Open Source Code | Yes | We release an implementation of our model as an open source project.2 2https://github.com/felixykliu/NAOMI |
| Open Datasets | Yes | The PEMS-SF traffic time series [30] data contains 267 training and 173 testing sequences of length 144... We generate 4000 training and 1000 test sequences of Billiards ball trajectories in a rectangular world using the simulator from [31]... The basketball tracking dataset contains the trajectories of professional basketball players on offense with 107,146 training and 13,845 test sequences. |
| Dataset Splits | Yes | The PEMS-SF traffic time series [30] data contains 267 training and 173 testing sequences of length 144... We generate 4000 training and 1000 test sequences of Billiards ball trajectories in a rectangular world using the simulator from [31]... The basketball tracking dataset contains the trajectories of professional basketball players on offense with 107,146 training and 13,845 test sequences. |
| Hardware Specification | No | The paper mentions 'See Appendix for implementation and training details.' but no appendix is provided in the document. The main text does not specify any particular hardware (e.g., CPU, GPU models, memory, or cloud instances) used for running the experiments. |
| Software Dependencies | No | The paper mentions 'See Appendix for implementation and training details.' but no appendix is provided in the document. The main text does not provide specific ancillary software details with version numbers (e.g., Python, PyTorch, or specific libraries). |
| Experiment Setup | No | The paper describes some general aspects of the training (e.g., 'For deterministic dynamics, we use the mean squared error as our loss', 'We randomly choose the number of steps to be masked'), and refers to an appendix for more details ('See Appendix for implementation and training details.'), but it does not include concrete hyperparameter values (e.g., learning rate, batch size, number of epochs) or detailed training configurations in the main text. |