Skip to main content


Fig. 4 | EURASIP Journal on Advances in Signal Processing

Fig. 4

From: Fast dictionary learning from incomplete data

Fig. 4

Atom spikiness (left) as well as recovery of the random low-rank component (middle) and random dictionary (right) of the corruption adapted versus the unadapted learning algorithms with varying atom support sizes. Two types of random erasure patterns leading to 36 and 64% corruption are used

Back to article page