SpeedBoost: Anytime Prediction with Uniform Near-Optimality

By Alexander Grubb, Drew Bagnell. 2012

96 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Graph HyperNetworks for Neural Architecture Search (2018, background): “This is formalized in anytime prediction, (Grubb & Bagnell, 2012) the setting in which for each test example x , there is non-deterministic computational budget B drawn from the joint distribution P ( x , B ) .”
Glance and Focus: a Dynamic Approach to Reducing Spatial Redundancy in Image Classification (2020, methodology): “We consider two settings to evaluate our method: (1) budgeted batch classification [20], where the network needs to classify a set of test samples within a given computational budget; (2) anytime prediction [13, 20], where the network can be forced to output a prediction at any given point in time.”
Data-driven exemplar model selection (2014, methodology): “In robotics, or in any other applications requiring time-bounded response with bounded computation [16], it is critical to select and rank detectors out of a large pool.”
Multi-Scale Dense Networks for Resource Efficient Image Classification (2017, methodology): “In the anytime prediction setting (Grubb & Bagnell, 2012), there is a finite computational budget B > 0 available for each test example x.”
Distillation-Based Training for Multi-Exit Architectures (2019, background): “Early examples of truly anytime classiﬁers were based on streaming nearest neighbors [41] or on classiﬁer ensembles such as decision trees [9], random forests [10], or boosting [13].”
The Greedy Miser: Learning under Test-time Budgets (2012, methodology): “(Simultaneous with this publication, Grubb & Bagnell (2012) also proposed a complementary approach to incorporate feature cost into gradient boosting.)”