Figure 1From: Profiled support vector machines for antisense oligonucleotide efficacy predictionIllustration of Gaussian-like profiles for the penalization factor and the ε-insensitive region in the P-SVR approach. In this case, we penalize harder the committed errors in the higher and lower efficacy regions. Additionally, the insensitive region becomes wider in medium AO efficacies, and thus few AOs will contribute to the cost function and, consequently, become support vectors. Only one additional parameter is introduced in the formulation, i.e. the width of the Gaussian profile, σ P .Back to article page