Incremental and Approximate Inference for Faster Occlusion-based Deep CNN Explanations

Summary: Occlusion-based CNN explanations reframed as incremental view maintenance; algebraic framework with materialized views and multi-query optimization to reuse computation across inferences. Krypton prototype (CPU/GPU) achieves up to 5× exact and 35× approximate speedups using novel approximate inference techniques that exploit CNN semantics. (summarized by gpt-5-nano on Feb 09 2026)

Paper ID: 5686
Venue: SIGMOD
Year: 2019
Pagerank: 7.9877991e-05
Overall Rank: 2,863 | 80.09%
DOI: 10.1145/3299869.3319874

Incoming Non-self Citations Over Time

Authors

Incoming Citations (Sorted by Pagerank)

Showing 13 of 13 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank
683	Cerebro: A Data System for Optimized Deep Learning Model Selection	2020	VLDB	0.00018195476
1,940	SliceLine: Fast, Linear-Algebra-based Slice Finding for ML Model Debugging	2021	SIGMOD	0.00010020173
2,839	VolcanoML: Speeding up End-to-End AutoML via Scalable Search Space Decomposition	2021	VLDB	8.0378978e-05
4,725	GeCo: Quality Counterfactual Explanations in Real Time	2021	VLDB	5.9697637e-05
4,774	LIMA: Fine-grained Lineage Tracing and Reuse in Machine Learning Systems	2021	SIGMOD	5.9316087e-05
6,903	POEM: Pattern-Oriented Explanations of Convolutional Neural Networks	2023	VLDB	4.8925595e-05
8,182	SHiFT: An Efficient, Flexible Search Engine for Transfer Learning	2023	VLDB	4.5659133e-05
8,864	Cerebro: A Layered Data Platform for Scalable Deep Learning	2021	CIDR	4.4326439e-05
9,192	Hyper-Tune: Towards Efficient Hyper-parameter Tuning at Scale	2022	VLDB	4.3765131e-05
11,431	Ease.ML: A Lifecycle Management System for MLDev and MLOps	2021	CIDR	4.1945683e-05
11,483	Shahin: Faster Algorithms for Generating Explanations for Multiple Predictions	2021	SIGMOD	4.1945683e-05
13,171	Reimagining Deep Learning Systems Through the Lens of Data Systems	2024	VLDB	-
13,313	Demonstration of Krypton: Optimized CNN Inference for Occlusion-based Deep CNN Explanations	2019	VLDB	-

Outgoing Citations (Sorted by Pagerank)

Showing 5 of 5 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
82	Answering Queries Using Views (Extended Abstract)	1995	PODS	0.00054402763
316	NoScope: Optimizing Neural Network Queries over Video at Scale	2017	VLDB	0.00027988668
449	Approximate Query Processing: Taming the TeraBytes! A Tutorial	2001	VLDB	0.00022846068
2,255	LINVIEW: Incremental View Maintenance for Complex Analytical Queries	2014	SIGMOD	9.1884983e-05
4,574	Incremental View Maintenance over Array Data	2017	SIGMOD	6.0738556e-05

Semantically Similar Papers

Overall Rank	Paper	Year	Venue	Pagerank
11,061	Optimizing Video Queries with Declarative Clues	2024	VLDB	4.1945683e-05
6,000	DeepEverest: Accelerating Declarative Top-K Queries for Deep Neural Network Interpretation	2022	VLDB	5.2415551e-05
2,886	VISTA: Optimized System for Declarative Feature Transfer from Deep CNNs at Scale	2020	SIGMOD	7.9612767e-05
9,764	View-based Explanations for Graph Neural Networks	2024	SIGMOD	4.2856106e-05
329	Accelerating Machine Learning Inference with Probabilistic Predicates	2018	SIGMOD	0.00027249545
9,120	Deep Query Optimization	2019	SIGMOD	4.392741e-05
13,229	Using Deep Learning Models to Replace Large Materialized Views in Relational Database	2021	CIDR	-
10,269	Database Views as Explanations for Relational Deep Learning	2026	VLDB	4.1945683e-05
3,293	Jointly Optimizing Preprocessing and Inference for DNN-based Visual Analytics	2021	VLDB	7.2629834e-05
13,313	Demonstration of Krypton: Optimized CNN Inference for Occlusion-based Deep CNN Explanations	2019	VLDB	-