RNN Step

Compute one step of a vanilla (Elman) RNN: the new hidden state is $h^{\mathrm{\prime }}=\tanh (W_{x}x+W_{h}h+b)h\text{'}=\backslash tanh(W\_xx+W\_hh+b)$ for the input vector x, previous hidden state h, input weights Wx, recurrent weights Wh and bias b.

Implement solve(x, h, Wx, Wh, b).

Returns. Return the new hidden state as a NumPy vector with entries in $(-1,1)(-1,1)$.

Examples

Example 1

Input:

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

Output:

[0.761594, 0.995055]

Example 2

Input:

solve([0.0], [0.0], [[1.0]], [[1.0]], [0.5])

Output:

[0.462117]