Dropout Mask

Apply inverted dropout to activations x during training. Draw a keep-mask with np.random.default_rng(seed).random(x.shape) < keep_prob, zero out the dropped entries, and divide the survivors by keep_prob so the expected value of each activation is unchanged.

Implement solve(x, keep_prob, seed=0).

Returns. Return a NumPy array with the shape of x. keep_prob must lie in $(0,1](0,1]$, otherwise ValueError is raised; keep_prob=1 returns x unchanged.

Examples

Example 1

Input:

solve([1.0, 2.0, 3.0, 4.0], 0.5, 0)

Output:

[0.0, 4.0, 6.0, 8.0]

Example 2

Input:

solve([1.0, 2.0, 3.0], 1.0, 0)

Output:

[1.0, 2.0, 3.0]

Example 3

Input:

solve([1.0, 2.0], 0.0, 0)

Output: Raises ValueError.