Figure 5.

Na_V regression results for all valid models. Efficiency is expressed in terms of prediction interval width at a fixed confidence of 0.8; smaller values are preferable. The colored areas correspond to the 95% CI computed from the ten replicate runs. The ICP_old^new was the only possible strategy to use when only having 30 available compounds in A_new but was surpassed in efficiency in all other experimental setups. CCP_AT and CCP_new have very similar efficiencies, and there is no clear winning strategy.