Forecasting Travel Speed in the Rainfall Days to Develop Suitable Variable Speed Limits Control Strategy for Less Driving Risk

Algorithm 1

CD-k algorithm for the RBMs.

	Input: Training data S = {x1, x2, xi}, the learning rate γ; number of steps k; the visiblelayer neuron: (); the hidden layer neuron: hj ()
Output: Connection weight matrix: W; Visual layer bias vector: b; Hidden layer bias vector: c.
1. Randomly initialize internal parameter vector: W, b, c.
2. For every traning sample xi in S,
for do
for , calculate
for , calculate
for ;
End