This is the training code for our paper "Diffusion Language-Shapelets for Semi-supervised Time-Series Classification" (AAAI-24).
Semi-supervised time-series classification could effectively alleviate the issue of lacking labeled data. However, existing approaches usually ignore model interpretability, making it difficult for humans to understand the principles behind the predictions of a model. Shapelets are a set of discriminative subsequences that show high interpretability in time series classification tasks. Shapelet learning-based methods have demonstrated promising classification performance. Unfortunately, without enough labeled data, the shapelets learned by existing methods are often poorly discriminative, and even dissimilar to any subsequence of the original time series. To address this issue, we propose the Diffusion Language-Shapelets model (DiffShape) for semi-supervised time series classification. In DiffShape, a self-supervised diffusion learning mechanism is designed, which uses real subsequences as a condition. This helps to increase the similarity between the learned shapelets and real subsequences by using a large amount of unlabeled data. Furthermore, we introduce a contrastive language-shapelets learning strategy that improves the discriminability of the learned shapelets by incorporating the natural language descriptions of the time series. Experiments have been conducted on the UCR time series archive, and the results reveal that the proposed DiffShape method achieves state-of-the-art performance and exhibits superior interpretability over baselines.
In accordance with the recommendation provided by the creators of the UCR archive and TS-TFC, we implemented a restriction to maintain an average of at least 30 samples per class within each dataset. This measure was taken to enhance the stability of the classification test results. As a result, we employed a total of 106 datasets from the initial pool of 128 UCR datasets for our experimental analysis.
To train a DiffShape model on a dataset for semi-supervised time series classification, run
python diffshape_ssc/main_diffshape.py --dataset [name of the dataset you want to train] ...Also, to train a DiffShape model on a dataset with few labeled samples per class for time series classification, run
python diffshape_ssc/main_diffshape_few.py --dataset [name of the dataset you want to train] ...