End-to-end learning of visual representations from uncurated instructional videos
Annotating videos is cumbersome, expensive and not scalable. Yet, many strong video
models still rely on manually annotated data. With the recent introduction of the HowTo100M …
models still rely on manually annotated data. With the recent introduction of the HowTo100M …
Weakly-supervised learning of visual relations
This paper introduces a novel approach for modeling visual relations between pairs of
objects. We call relation a triplet of the form (subject, predicate, object) where the predicate …
objects. We call relation a triplet of the form (subject, predicate, object) where the predicate …
On adaptive sketch-and-project for solving linear systems
We generalize the concept of adaptive sampling rules to the sketch-and-project method for
solving linear systems. Analyzing adaptive sampling rules in the sketch-and-project setting …
solving linear systems. Analyzing adaptive sampling rules in the sketch-and-project setting …
Optimal margin distribution machine
Support Vector Machine (SVM) has always been one of the most successful learning
algorithms, with the central idea of maximizing the minimum margin, ie, the smallest distance …
algorithms, with the central idea of maximizing the minimum margin, ie, the smallest distance …
Safe adaptive importance sampling
Importance sampling has become an indispensable strategy to speed up optimization
algorithms for large-scale applications. Improved adaptive variants--using importance …
algorithms for large-scale applications. Improved adaptive variants--using importance …
A flexible model for training action localization with varying levels of supervision
Spatio-temporal action detection in videos is typically addressed in a fully-supervised setup
with manual annotation of training videos required at every frame. Since such annotation is …
with manual annotation of training videos required at every frame. Since such annotation is …
Faster coordinate descent via adaptive importance sampling
D Perekrestenko, V Cevher… - Artificial Intelligence and …, 2017 - proceedings.mlr.press
Coordinate descent methods employ random partial updates of decision variables in order
to solve huge-scale convex optimization problems. In this work, we introduce new adaptive …
to solve huge-scale convex optimization problems. In this work, we introduce new adaptive …
On Frank-Wolfe and equilibrium computation
JD Abernethy, JK Wang - Advances in Neural Information …, 2017 - proceedings.neurips.cc
Abstract We consider the Frank-Wolfe (FW) method for constrained convex optimization, and
we show that this classical technique can be interpreted from a different perspective: FW …
we show that this classical technique can be interpreted from a different perspective: FW …
Lazifying conditional gradient algorithms
G Braun, S Pokutta, D Zink - Journal of Machine Learning Research, 2019 - jmlr.org
Conditional gradient algorithms (also often called Frank-Wolfe algorithms) are popular due
to their simplicity of only requiring a linear optimization oracle and more recently they also …
to their simplicity of only requiring a linear optimization oracle and more recently they also …
Restarting frank-wolfe
T Kerdreux, A d'Aspremont… - The 22nd international …, 2019 - proceedings.mlr.press
Abstract Conditional Gradients (aka Frank-Wolfe algorithms) form a classical set of methods
for constrained smooth convex minimization due to their simplicity, the absence of projection …
for constrained smooth convex minimization due to their simplicity, the absence of projection …