Multi-label literature classification based on the Gene Ontology graph

BMC Bioinformatics

Table 2 Algorithm 2 Random GO walk

1	inputs:
	d, a new document
	G, a GO graph with training cases
	nSample, the sample size
	nMaxSteps, the number of maximum steps
2	initialize:
3	finalLeaves ← {}
4	finalLeavesProbs ← {}
5	for n in 1: nSamples
6	g ← initialize()
7	//Select initial GO node randomly
	for s in 1: nMaxSteps:
8	T ← TempFunc(s)
9	nbrs ← G.neighbors(g)
10	g* ← q(g\|g, d*)
	//Sample from proposal distribution, g∈{g, nbrs*}
11	u ← uniform [0, 1]
12	if u <A ← $\min (1, \frac{p {(g^{} \| d)}^{\frac{1}{T}} q (g \| g^{}, d)}{p {(g \| d)}^{\frac{1}{T}} q (g^{*} \| g, d)})$
13	g ← g*
14	end
15	finalLeaves ← union(finalLeaves, curNode)
16	end
17	finaLeavesProbs ← calProbFromSample(finalLeaves)
18	outputs: finalLeaves, finaLeavesProbs

ISSN: 1471-2105