Algorithm 2 Compressed learning memory selection.

Input: Compressed samples $X1$, Non-Compressed samples $X2$, Number of samples per class $N_t$, Compression rate p, Total number of seen classes $C_t$; Initialize: Memory M← {} * M; 1:for<$c=0,1,2,..$$C_t$>do                                                       ▹ Iterate through all seen classes. 2:    nc$,$X1c$,$X2c ←Nt[c]$,$X1[c]$,$X2[c]     ▹ Populate data from the compressed learning process. 3:    $\mathit{M}\leftarrow \mathit{M}+\mathit{RandSelect}(X{1}_c,\mathit{min}$(nc, len(X1c)))     ▹ Random sampling through the Compressed samples. 4:    nc$\leftarrow (n_c-\mathit{min}$(nc, len(X1c))$) \ast p$▹ Update the maximum number of available spaces for the Non-Compressed samples. 5:    $\mathit{M}\leftarrow \mathit{M}+\mathit{RandSelect}(X{2}_c,\mathit{min}$(nc, len(X2c)))         ▹ Random sampling through the Non-Compressed samples.