Embedded Files
Current Openings
Deep Sparse Matrix Factorization (DSMF):
Features derived using sparse representation (SR) based dictionary learning approaches have been shown to yield promising results for various signal processing tasks. In practice such single-level decomposition may not be suitable for the signals having complex hierarchical information about various hidden attributes. Hence, in DSMF the dictionary is further factorized in a way such that each factor adds an extra layer of abstraction.
Applications to
Speech Recognition
Different Speech Units Have More Discriminative Representation At One Layer As Compared To Other Layers.
Functional Connectivity Analysis in Resting State fMRI
Convectional Model Order: 60 Model Order (Proposed): 20
Max Networks Reported: 10-17 Networks Reported (Proposed): 17
1: Default Mode Network (aDMN), 2: Posterior Default Mode Network (pMDN), 3: Dorsal Attention Network (DAN), 4: Primary Attention Network (PAN), 5: Right Fronto Parietal Executive Network (rFPE), 6: Left Fronto-Parietal Executive Network (lFPE), 7: Primary Executive Network (ExP), 8: Central Executive Network (CEN), 9: Sensory Motor Network (SMN), 10: Motor Network, 11: Auditory (Aud), 12: Salience Network (Sal), 13: Ventral Stream (VS), 14: Primary Visual (VisP), 15: Medial Visual (VisM), 16: Lateral Visual (VisL), 17: Occipital Visual (VisO).
Pseudo Activations Suppressed in Deeper Layers
Bird Species Identification
Sparse-Dense-Sparse Processing For Extracting Audio Embedding
Page updated
Google Sites
Report abuse