Data Imputation for Detected Traffic Volume of Freeway Using Regression of Multilayer Perceptron

Algorithm 1

L-BFGS.

	Definition:
	θ0: the first run parameter of MLP
	μ: the tolerable maximum positive number
	iter: iteration of L-BFGS, iter = k(k = 0,1,2, …)
	m: latest m groups of iteration results are used for calculation
	ɛ: optimized step length
	Procedure L-BFGS
	Calculate B0 and according to θ0 and the value of the loss function
	get (θk + ɛr) = min(f(θk + ɛir)) from:
	Let:
	for i = k − 1, k − 2, …, k − m:
	end for
	r = B·
	for i = k − m, k − m + 1, …, k − 1:
	end for
	k = k + 1
	end while
	end procedure