← All models
architecture

Temporal Convolutional Network

intermediatesupervised · self_supervised
parametricclassificationregressionforecastingrepresentationtime_seriessequencesaudio

A TCN uses causal, usually dilated one-dimensional convolutions to model sequences with a large and controllable history.

Mechanisms

causal convolutiondilationbackpropagation

Properties

nonlinearrepresentation learning

Constraints

requires large datarequires scalingsensitive to tuning

Practical profile

Explainability
low
Training cost
high
Inference cost
medium
Data appetite
high