Anchor Attention for Hybrid Crowd Forecasts Aggregation.
Forecasting the future is a notoriously difficult task. One way to address this challenge is to "hybridize" the forecasting process, combining forecasts from a crowd of humans, as well as one or more machine models. However, an open challenge remains in how to optimally aggregate inputs from these pools into a single forecast. We proposed anchor attention for this type of sequence summary problem. Each forecast is represented by a trainable embedding vector. An anchor attention score is used to determine input weights. We evaluate our approach using data from a real-world forecasting tournament, and show that our method outperforms the current state-of-the-art aggregation approaches.