Attention Augmented Convolutional Transformer for Tabular Time-series
This work addresses time-series classification in industrial settings, offering a scalable solution with novel techniques for pretraining and feature handling, though it appears incremental by building on existing transformer and convolution methods.
The authors tackled the problem of time-series classification for industrial tabular data by proposing a novel scalable architecture that integrates one-dimensional convolutions with transformers and introduces a new masking technique for pretraining, achieving an end-to-end representation learning framework that handles both categorical and continuous inputs without quantization.
Time-series classification is one of the most frequently performed tasks in industrial data science, and one of the most widely used data representation in the industrial setting is tabular representation. In this work, we propose a novel scalable architecture for learning representations from tabular time-series data and subsequently performing downstream tasks such as time-series classification. The representation learning framework is end-to-end, akin to bidirectional encoder representations from transformers (BERT) in language modeling, however, we introduce novel masking technique suitable for pretraining of time-series data. Additionally, we also use one-dimensional convolutions augmented with transformers and explore their effectiveness, since the time-series datasets lend themselves naturally for one-dimensional convolutions. We also propose a novel timestamp embedding technique, which helps in handling both periodic cycles at different time granularity levels, and aperiodic trends present in the time-series data. Our proposed model is end-to-end and can handle both categorical and continuous valued inputs, and does not require any quantization or encoding of continuous features.