> ML_ALGORITHM // INFORMER-PROBSPARSE-ATTENTION_v1.0
Informer (ProbSparse Attention)
Efficient Transformer for long-sequence time-series forecasting that reduces quadratic self-attention complexity to O(L log L) via ProbSparse attention.
Transformer-Based Forecastingtime-series-forecastingmoderate-posthoclarge (>100k)
Back to All AlgorithmsComputational Complexity
Training Complexity:O(L * log(L) * d)
Inference Complexity:O(L * log(L) * d)
Hardware Profile
CPU Friendly:No
Requires GPU:Yes
Memory Footprint:moderate
Interpretability & Data
Interpretability Tier:moderate-posthoc
Training Data Needs:large (>100k)
Interpretability Assessment
Reduces classical quadratic Transformer memory to O(L log L) via selective active queries.
Suitable Tasks & Supported Modalities
Suitable Tasks:
long horizon forecastingmultivariate time series forecasting
Supported Modalities:
time-series
Implementing Libraries
Foundational Literature
Common Pitfalls & Warnings
- On many standard benchmarks, simple single-layer linear models (DLinear) outperform Informer with 1/100th the compute (the "Are Transformers Effective for Time Series?" critique)