Topologically Ordered Feature Extraction Based on Sparse Group Restricted Boltzmann Machines

Figure 5

(a) Activation probabilities computed under the SRBMs, the sparseness of the vector is 0.68; (b) activation probabilities computed under the SGRBMs; the sparseness is 0.89; (c) activation probabilities computed under the NTOSGRBMs; the sparseness is 0.81; (d) activation probabilities computed under the TOSGRBMs; the sparseness is 0.78.

(a) SRBMs

(b) SGRBMs

(c) NTOSGRBMs

(d) TOSGRBMs