DeepGPD: A Deep Learning Approach for Modeling Geospatio-Temporal Extreme Events
Authors: Tyler Wilson, Pang-Ning Tan, Lifeng Luo4245-4253
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the effectiveness of our approach on a realworld dataset for modeling extreme climate events. |
| Researcher Affiliation | Academia | Tyler Wilson,1 Pang-Ning Tan,1 Lifeng Luo2 1 Dept of Computer Science and Engineering, Michigan State University 2 Department of Geography, Michigan State University {wils1270,ptan,lluo}@msu.edu |
| Pseudocode | No | The paper describes the model architecture and algorithms in text and diagrams, but it does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code and data for our implementation is available at https: //github.com/Tyler PWilson/deep GPD. |
| Open Datasets | Yes | We evaluate our proposed framework on a 44-year global precipitation data from 1970 to 2013. Specifically, we use daily precipitation values collected from the Global Historical Climatology Network1 (GHCN) for 1,112 stations located in the Northern Hemisphere (between 22.5 N to 67.5 N) as our target variable. ... 1https://www.ncdc.noaa.gov/ghcn-daily-description |
| Dataset Splits | Yes | Each 1 year window of predictor and target values are assigned to either training, validation or test sets, with 34 windows in training, 5 in validation and 4 in test. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments (e.g., specific GPU or CPU models, memory, or cluster specifications). |
| Software Dependencies | No | The paper mentions using 'Adam (Kingma and Ba 2015)' as an optimizer but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | Hyper-parameters for all deep learning models were selected as follows: learning rates between 10 2 and 10 5, number of layers ranged from 3 to 10, and hidden dimensions between 5 and 50 units were explored. |