Temporal Convolutional Networks
Also known as TCN
Overview
Temporal Convolutional Networks (TCNs) apply one-dimensional convolutions across time to recognize patterns in sequential data, often using causal, dilated, and residual layers to cover long histories efficiently. They are used for forecasting and classification in settings such as traffic, wearable sensors, EEG, speech, motion, and clinical signals, where parallel convolution can be a practical alternative to recurrent models. Common extensions add multi-scale filters, attention, graph structure, or physics-informed constraints to capture different temporal scales and domain relationships. TCNs are competitive in many tasks, but longer-range dependencies, robustness across subjects and settings, and whether strong predictions reflect useful temporal learning remain important challenges.
Latest papers
31 papers2 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.