Optimal Parallelization of Boosting
This work addresses the parallel complexity of Boosting for machine learning practitioners, settling the true parallel complexity for nearly sample-optimal algorithms.
The paper tackles the problem of parallelizing Boosting algorithms by closing the gap between theoretical lower bounds and existing algorithms, providing improved lower bounds and a new algorithm that matches these bounds across the entire tradeoff spectrum up to logarithmic factors.
Recent works on the parallel complexity of Boosting have established strong lower bounds on the tradeoff between the number of training rounds $p$ and the total parallel work per round $t$. These works have also presented highly non-trivial parallel algorithms that shed light on different regions of this tradeoff. Despite these advancements, a significant gap persists between the theoretical lower bounds and the performance of these algorithms across much of the tradeoff space. In this work, we essentially close this gap by providing both improved lower bounds on the parallel complexity of weak-to-strong learners, and a parallel Boosting algorithm whose performance matches these bounds across the entire $p$ vs.~$t$ compromise spectrum, up to logarithmic factors. Ultimately, this work settles the true parallel complexity of Boosting algorithms that are nearly sample-optimal.