Whatever it takes to prove that √(log k/ε² + log log n) error bound with 95% probability – “probably almost correct” – what else do you need, really?
Whatever it takes to prove that √(log k/ε² + log log n) error bound with 95% probability – “probably almost correct” – what else do you need, really?
In my experience it runs much smoother on proton than the native version