An Implicit Memory-Based Method for Supervised Pattern Recognition

Figure 9

Prediction result by a base neural network on a training image with a label of (a) 0, (b) 1, (c) 2, (d) 3, (e) 4, (f) 5, (g) 6, (h) 7, (i) 8, and (j) 9 using randomly generated masks. The mask covers up a rectangular part of the image, and the predicted result is in the box.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)