A Probabilistic Re-Intepretation of Confidence Scores in Multi-Exit Models
In this paper, we propose a new approach to train a deep neural network with multiple intermediate auxiliary classifiers, branching from it. These ‘multi-exits’ models can be used to reduce the inference time by performing early exit on the intermediate branches, if the confidence of the prediction...
Main Authors: | Jary Pomponi, Simone Scardapane, Aurelio Uncini |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2021-12-01
|
Series: | Entropy |
Subjects: | |
Online Access: | https://www.mdpi.com/1099-4300/24/1/1 |
Similar Items
-
Depth-Adaptive Deep Neural Network Based on Learning Layer Relevance Weights
by: Arwa Alturki, et al.
Published: (2022-12-01) -
Sampling the Variational Posterior with Local Refinement
by: Marton Havasi, et al.
Published: (2021-11-01) -
Statistical inference in generative models using scoring rules
by: Pacchiardi, L
Published: (2022) -
Probabilistic Models with Deep Neural Networks
by: Andrés R. Masegosa, et al.
Published: (2021-01-01) -
Confidence Score: The Forgotten Dimension of Object Detection Performance Evaluation
by: Simon Wenkel, et al.
Published: (2021-06-01)