“Functional Gradient Descent with Adaptive Representations” accepted at NeurIPS
The announcement claims the work solves a practical hurdle: accurately implementing functional gradient descent algorithms.
TLDR
A post announcing the paper’s acceptance at NeurIPS says functional gradient descent algorithms generally outperform neural networks but are hard to implement accurately. It claims the new work fixes that implementation problem.
Combined views
29
1 Source, first seen 4h ago
41 reposts
