What are the three components of the temporal feature extractor in PPGF, and what role does each play?

The three components are a 1D convolutional layer, a Transformer encoder, and a gated residual network (GRN). The convolutional layer extracts local temporal features and short-range dependencies, the Transformer encoder captures long-range interactions via self-attention, and the GRN adaptively modulates the feature representations to handle variability and complexity in the time series.

The temporal feature extractor in PPGF is a hierarchical encoder with three components. First, a 1D convolutional layer rapidly extracts local temporal features, effectively capturing short-range dependencies and localized patterns in the input sequence. Second, a Transformer encoder is applied after convolution to learn long-range interactions; its self-attention mechanism enhances the model's ability to capture temporal dependencies across broader horizons, complementing the convolutional layer's local focus. Third, a gated residual network (GRN) adaptively modulates feature representations, and its gating structure supports dynamic depth control, making it well suited for managing the variability and complexity of multivariate time series. Together, these components enable the model to learn both localized structures and intricate long-range patterns, producing a general temporal representation g for each input rather than relying solely on fixed receptive fields.