WPMixer: Efficient Multi-Resolution Mixing for Long-Term Time Series Forecasting

Authors: Md Mahmuddun Nabi Murad, Mehmet Aktukmak, Yasin Yilmaz

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We extensively evaluate the long-term forecasting performance of WPMixer on 7 popular datasets: ETTh1, ETTh2, ETTm1, ETTm2, Weather, Electricity, and Traffic. ... We conducted an extensive ablation study to evaluate the individual contribution of each module within the proposed model using the ETT datasets.
Researcher Affiliation Collaboration Md Mahmuddun Nabi Murad1, Mehmet Aktukmak2, Yasin Yilmaz1 1University of South Florida 2Intel Corporation EMAIL, EMAIL, EMAIL
Pseudocode No The paper describes the model architecture and components in detail through descriptive text and mathematical equations, but does not include a distinct, structured pseudocode or algorithm block.
Open Source Code Yes Code is available at https://github.com/Secure-and-Intelligent-Systems-Lab
Open Datasets Yes We extensively evaluate the long-term forecasting performance of WPMixer on 7 popular datasets: ETTh1, ETTh2, ETTm1, ETTm2, Weather, Electricity, and Traffic. The specifications of datasets are given in Table 1.
Dataset Splits Yes The specifications of datasets are given in Table 1. Dataset size refers to the training, validation, and testing dataset sizes. ... Following the practice in Informer, Autoformer, Patch TST, TSMixer, and Time Mixer, all datasets were normalized to a zero mean and unit standard deviation. The normalized datasets served as the basis for ground truth in our evaluations.
Hardware Specification Yes Experiments with the ETT and Weather datasets were performed on a single NVIDIA Ge Force RTX 4090 GPU while the experiments with the Electricity and Traffic datasets were carried out using two NVIDIA A100 GPUs.
Software Dependencies No The paper mentions 'PyTorch' and 'Optuna' but does not provide specific version numbers for these software components or any other libraries.
Experiment Setup Yes In long-term forecasting, the lengths of predictions were set at 96, 192, 336, and 720, in alignment with prior studies. During the training phase, Smooth L1Loss was employed, whereas Mean Squared Error (MSE) and Mean Absolute Error (MAE) were utilized for evaluation purposes. ... We used Optuna (Akiba et al. 2019) with the default setting of Tree-structured Parzen Estimator (TPE) for optimizing the hyperparameters. The optimized hyperparameter values are shown in Table 7 in (Murad, Aktukmak, and Yilmaz 2024). ... The other parameters are kept fixed for all m as follows, look-back window 512, initial learning rate 0.001, wavelet type Daubechies 5, batch size 128, epochs 10, d = 256, tf = 7, df = 7, patch size 16, and stride 8.