Sparsity Increases Uncertainty Estimation in Deep Ensemble

Deep neural networks have achieved almost human-level results in various tasks and become popular the broad artificial intelligence domains. Uncertainty estimation is an on-demand task caused by black-box point behavior of deep learning. The ensemble provides increased accuracy estimated uncertainty; however, linearly increasing size makes unfeasible for memory-intensive tasks. To address this problem, we used model pruning quantization with a analyzed effect context uncertainty metrics. We empirically showed that members’ disagreement increases pruning, making models sparser zeroing irrelevant parameters. Increased im-plies uncertainty, which helps more robust predictions. Accordingly, energy-efficient compressed appropriate uncertainty-aware