Source code and experimental results for the paper "A Deep Dive into Alternatives to the Global Average Pooling for Time Series Classification" AALTD 2025.
A study on candidates to replace Global Average Pooling (GAP) in neural network architectures for time series classification.
- Download and unpack the 2018 UCR archive 📎
- Download and unpack the 2018 UAE multivariate TSC
📎
- The CharacterTrajectories dataset in the archive causes problems (metadata related), to solve them, download the CharacterTrajectories 📎 dataset and replace with the original one.
- Change
data_folderindata.pyaccording to the previous steps.
- Python
3.11 - TensorFlow
2.16.1- Deep learning
- NumPy
1.26.4- Data manipulation
- aeon
1.0- Read
.tsdataset files
- Read
- scikit-learn
1.5.2- Preprocess labels