Algorithm 1.

CSC Algorithm

Input: Training set $\mathbf{X}={\{{\mathbf{x}}_i\}}_{i=1}^N$, K, α
Output: Convolutional filter bank $\mathbf{D}={\{{\mathbf{d}}_k\}}_{k=1}^K$
1:	Initialize: D ∼ (0, 1), Z ← 0
2:	Repeat
3:	for i = 1 to N do
4:	Normalize each kernel in D to unit energy
5:	Fixing D, compute sparse feature maps Zi by solving $${\mathbf{Z}}^i\leftarrow arg\underset{{\mathbf{z}}_k^i\in {\mathbf{Z}}^i}{min}{\Vert {\mathbf{x}}_i-{\displaystyle \sum _{k=1}^K}{\mathbf{d}}_k\ast {\mathbf{z}}_k^i\Vert }_{\mathrm{F}}^2+\alpha {\displaystyle \sum _{k=1}^K}{\Vert {\mathbf{z}}_k^i\Vert }_1$$
6:	Fixing Z, update D as
	D ← D − μ ∇D (D, Z)
7:	end for
8:	until Convergence (maximum iterations reached or objective function ≤ threshold)