Implements a transformer architecture inspired by TCDformer for time series changepoint detection using self-attention mechanisms.
transformer_detect(
data,
true_changepoints = NULL,
window_size = 128,
d_model = 64,
n_heads = 4,
n_layers = 2,
d_ff = 256,
dropout = 0.1,
epochs = 50,
threshold = 0.5,
verbose = FALSE
)List with changepoints, attention weights, and model
Numeric vector of time series data
Optional vector of known changepoints
Input window size (default: 128)
Model dimension (default: 64)
Number of attention heads (default: 4)
Number of transformer layers (default: 2)
Feed-forward dimension (default: 256)
Dropout rate (default: 0.1)
Training epochs (default: 50)
Detection threshold (default: 0.5)
Show progress (default: FALSE)
Wu, H., et al. (2023). TimesNet: Temporal 2D-Variation Modeling
Zhou, H., et al. (2021). Informer: Efficient Transformer for Long Sequence Time-Series Forecasting