Spaces:
Paused
Paused
xiv | |
CONTENTS | |
2.3 | |
2.4 | |
I | |
2.2.4 Calibration . . . . . . . . . . . . . . . . | |
2.2.5 Predictive Uncertainty Quantification . | |
2.2.6 Failure Prediction . . . . . . . . . . . . | |
Document Understanding . . . . . . . . . . . . | |
2.3.1 Task Definitions . . . . . . . . . . . . . | |
2.3.2 Datasets . . . . . . . . . . . . . . . . . . | |
2.3.3 Models . . . . . . . . . . . . . . . . . . | |
2.3.4 Challenges in Document Understanding | |
2.3.4.1 Long-Context Modeling . . . . | |
2.3.4.2 Document Structure Modeling | |
Intelligent Automation . . . . . . . . . . . . . . | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
. | |
Reliable and Robust Deep Learning | |
3 Benchmarking Scalable Predictive Uncertainty in Text Classification | |
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
3.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
3.3 Uncertainty Methods . . . . . . . . . . . . . . . . . . . . . . . . | |
3.3.1 Quantifying Uncertainty in Deep Learning . . . . . . . . | |
3.3.2 Predictive Uncertainty Methods . . . . . . . . . . . . . | |
3.3.2.1 Monte Carlo Dropout . . . . . . . . . . . . . . | |
3.3.2.2 Deep Ensemble . . . . . . . . . . . . . . . . . . | |
3.3.2.3 Concrete Dropout . . . . . . . . . . . . . . . . | |
3.3.2.4 Heteroscedastic Extensions . . . . . . . . . . . | |
3.3.3 Uncertainty Estimation . . . . . . . . . . . . . . . . . . | |
3.3.4 Motivating Hybrid Approaches . . . . . . . . . . . . . . | |
3.3.5 Uncertainty Calibration under Distribution Shift . . . . | |
3.4 Experimental Methodology . . . . . . . . . . . . . . . . . . . . | |
3.4.1 Proposed Hybrid Approaches . . . . . . . . . . . . . . . | |
3.4.2 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
3.4.3 Architecture . . . . . . . . . . . . . . . . . . . . . . . . | |
3.4.4 Evaluation metrics . . . . . . . . . . . . . . . . . . . . . | |
3.4.5 Experimental design . . . . . . . . . . . . . . . . . . . . | |
3.4.5.1 In-domain Setting . . . . . . . . . . . . . . . . | |
3.4.5.2 Cross-domain Setting . . . . . . . . . . . . . . | |
3.4.5.3 Novelty Detection Setting . . . . . . . . . . . . | |
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
3.5.1 Experiment: In-domain . . . . . . . . . . . . . . . . . . | |
3.5.2 Experiment: Cross-domain . . . . . . . . . . . . . . . . | |
3.5.3 Experiment: Novelty Detection . . . . . . . . . . . . . . | |
3.5.4 Experiment: Ablations . . . . . . . . . . . . . . . . . . . | |
3.5.4.1 Diversity . . . . . . . . . . . . . . . . . . . . . | |
28 | |
30 | |
32 | |
33 | |
35 | |
36 | |
37 | |
38 | |
39 | |
40 | |
41 | |
43 | |
44 | |
46 | |
48 | |
51 | |
51 | |
52 | |
53 | |
53 | |
54 | |
54 | |
55 | |
58 | |
59 | |
61 | |
61 | |
63 | |
64 | |
66 | |
66 | |
67 | |
67 | |
68 | |
69 | |
70 | |
71 | |
73 | |
75 | |
76 | |