Fig. 5.
Illustration of the D-r1DL framework. (a) Running example showing the input data S (one volume from the 4-D volumetric matrix), learned vector v (3-D volumetric matrix as a vector) and vector u (time series). (b) Algorithmic pipeline of r1DL. Red arrow shows the updating loop for learning each [u, v], blue arrow shows the updating loop for deflation of S and learning next dictionary. (c) Parallelization steps for the three operations from (b).