Difficulty: Easy Tags: Optimization
NVIDIA's internal GPU-scheduler team tunes a lightweight utilization-prediction model with plain SGD before ever touching anything fancier. Every optimizer in the more advanced tracks builds on getting this single update exactly right.
theta <- theta - eta * grad
d eta
theta_1 ... theta_d
grad_1 ... grad_d
Updated theta, d values, 6 decimals.
1 <= d <= 10^4Input
2 0.1
1.0 2.0
0.2 -0.1
Output
0.980000 2.010000
No results yet
Run your code to check it against the examples, or submit to run every test.