SC2 participant models were aggregated using two types of ensemble models Spectral Meta-Learner (SML) and Random Aggregation. SML choses predictions from n methods to aggregate based on an estimation of BAC for each method without using actual labels15,53. Random Aggregation is the traditional way that people aggregate models by giving equal weight to each method. We randomly pick n methods (do this ten times) and for n methods we compute the average BAC and the error.
notion d ensemble pdf download
Download: https://vittuv.com/2vIPHH
2ff7e9595c
Comments