3 ms·
Were other problems less about exploration? If not, I'd argue lasertag might be more important. In various benchmarks the geometric mean is usually used to com
by lostmsu 6y ago
Were other problems less about exploration? If not, I'd argue lasertag might be more important.
In various benchmarks the geometric mean is usually used to compare total score across different tasks to account for severe issues with specific tasks.
If your network is the same as IMPALA, why do you show its results with different hyperparameters? Were some of them necessary for the optimization (e.g. reduced batch size)?