Learning Optimal Projection for Forecast Reconciliation of Hierarchical Time Series
Authors: Asterios Tsiourvas, Wei Sun, Georgia Perakis, Pin-Yu Chen, Yada Zhu
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | An empirical evaluation of real-world hierarchical time series datasets demonstrates the superior performance of the proposed method over existing state-of-the-art approaches. |
| Researcher Affiliation | Collaboration | 1Operations Research Center, Massachusetts Institute of Technology, USA 2IBM Research, USA. |
| Pseudocode | No | The paper describes methods and processes through textual explanation and mathematical formulations, but it does not include a clearly labeled 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | No | The paper does not explicitly state that the source code for the described methodology is being released or provide a direct link to its repository. |
| Open Datasets | Yes | We evaluate our proposed methodology on the publicly available hierarchical datasets used in Rangapuram et al. (2021). We consider the Labour dataset (Australian Bureau of Statistics, 2020)... the Traffic dataset (Cuturi, 2011)... and the Wiki dataset (Ben Taieb & Koo, 2019)... Tourism dataset (Tourism Research Australia, 2005)... and on the Tourism Large dataset (Wickramasuriya et al., 2019)... |
| Dataset Splits | Yes | For performing cross-validation on the neural forecaster, we train on the first T − h time steps and validate on the following h time steps. |
| Hardware Specification | No | The paper does not explicitly provide specific hardware details such as GPU or CPU models, memory specifications, or cloud computing instance types used for running the experiments. |
| Software Dependencies | No | The paper mentions software like Gluon TS, Times Net, and Autoformer, and refers to libraries, but it does not specify exact version numbers for these software dependencies (e.g., 'PyTorch 1.9', 'Python 3.8'). |
| Experiment Setup | Yes | For both models, for the family of general oblique projection, we use a Lagrange multiplier of λ = 10^4, since this value is large enough to guarantee reconciliation. |