Mechanism diagram · CNN / TCN
CNN / TCN (DILATED) Input series lookback × N Conv 1×k dilation 1 Conv 1×k dilation 2 Conv 1×k dilation 4 Residual + LayerNorm stacked depth-wise blocks Pool + head Forecast

How it works

TCN stacks 1-D causal convolutions with increasing dilation factors.

Receptive field grows exponentially with depth: a TCN with N layers and base kernel size k has receptive field roughly k × 2^N.

Residual connections stabilise deep stacks.

Pros and cons on this universe

Pros

  • Parallelisable across time — much faster than RNN at training.
  • Standard baseline in the time-series literature.
  • Receptive field is interpretable and easy to tune.

Cons / failure modes

  • Rank #22 IC — weak on this universe.
  • Beaten by BITCN (bidirectional variant) and ModernTCN on the literature benchmark.
  • No dossier in the research base.

References