update even if the grad is zero, but in this sparse implementation, ms
and mom will not update in iterations during which the grad is zero.

mean_square = decay * mean_square + (1-decay) * gradient ** 2
Delta = learning_rate * gradient / sqrt(mean_square + epsilon)

ms <- rho * ms_{t-1} + (1-rho) * grad * grad
mom <- momentum * mom_{t-1} + lr * grad / sqrt(ms + epsilon)
var <- var - mom
  }];

## Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

4d55397500, Abe, Alistair Low, Andy Kernahan, Appledore, Ben, Ben Barsdell,
Boris Pfahringer, Brad Wannow, Brett Koonce, Carl Thomé, cclauss, Chengzhi Chen,
Chris Drake, Christopher Yeh, Clayne Robison, Codrut Grosu, Daniel Trebbien,
Danny Goodman, David Goodwin, David Norman, Deron Eriksson, Donggeon Lim, Donny

33AC          ; mapped                 ; 0067 0070 0061 #1.1  SQUARE GPA
33AD          ; mapped                 ; 0072 0061 0064 #1.1  SQUARE RAD
33AE          ; mapped                 ; 0072 0061 0064 2215 0073 #1.1 SQUARE RAD OVER S
33AF          ; mapped                 ; 0072 0061 0064 2215 0073 0032 #1.1 SQUARE RAD OVER S SQUARED
33B0          ; mapped                 ; 0070 0073     # 1.1  SQUARE PS