Quantum Inspired Computational Intelligence

4.3 Sparse Solutions

Another key feature in learning is the notion of sparsity. Sparsity allows one, for example, to perform an automatic detection of the most relevant features of the input space $\mathcal{X}$ in the set of the d possible ones for the particular learning task [135]. In other situations, sparsity is required for computational reasons, in the sense that sparse models can be deployed with reduced computational effort [76, 124, 136]. Finally, previous studies [24, 76] showed that increasing the sparsity of the solution leads to a beneficial regularization effect, allowing simpler models to be learned which outperform more complex ones while ...