Algorithm 2.1 Unlabelled Conformal Predictor in search of interactions
Input: training examples (known interactions) x1, x2,…, xl ∈ X
Input: a new example xl+1 ∈ X
Input: unlabelled examples U = {xl+2, xr+2,…, xl+s+1} ⊂ X
Input: a confidence level γ
Input: a NCM A
for j in 1, 2,…, l + 1 do
αj = A(xj, {x1,…, xl, xl+1},U)
end for $$p=\frac{\#\{j=1,\dots ,l+1:{\alpha }_j\ge {\alpha }_{l+1}\}}{l+1}$$
Output: xl+1 is predicted as a suspected interaction if p > 1 − γ.