The authors report that reducing FastGCN's early-stopping window to 10 epochs improved its performance and allowed shorter training time; they caution that FastGCN's originally published CPU benchmark is difficult to compare numerically with their GPU benchmark. · CiteArk
The authors report that reducing FastGCN's early-stopping window to 10 epochs improved its performance and allowed shorter training time; they caution that FastGCN's originally published CPU benchmark is difficult to compare numerically with their GPU benchmark.
