Capped -Norm Proximal Support Vector Machine

Figure 6

Training samples from MNIST  data. (a) Original training data. (b) Training data with 30% area black block noise. (c) Training data with 30% area Gaussian block noise.

(a)

(b)

(c)