Results 1 
7 of
7
Empirical Bernstein Stopping
"... Sampling is a popular way of scaling up machine learning algorithms to large datasets. The question often is how many samples are needed. Adaptive stopping algorithms monitor the performance in an online fashion and they can stop early, saving valuable resources. We consider problems where probabili ..."
Abstract

Cited by 65 (8 self)
 Add to MetaCart
(Show Context)
Sampling is a popular way of scaling up machine learning algorithms to large datasets. The question often is how many samples are needed. Adaptive stopping algorithms monitor the performance in an online fashion and they can stop early, saving valuable resources. We consider problems where probabilistic guarantees are desired and demonstrate how recentlyintroduced empirical Bernstein bounds can be used to design stopping rules that are efficient. We provide upper bounds on the sample complexity of the new rules, as well as empirical results on model selection and boosting in the filtering setting. 1.
Structure Learning of Bayesian Networks using Constraints
"... This paper addresses exact learning of Bayesian network structure from data and expert’s knowledge based on score functions that are decomposable. First, it describes useful properties that strongly reduce the time and memory costs of many known methods such as hillclimbing, dynamic programming and ..."
Abstract

Cited by 51 (6 self)
 Add to MetaCart
(Show Context)
This paper addresses exact learning of Bayesian network structure from data and expert’s knowledge based on score functions that are decomposable. First, it describes useful properties that strongly reduce the time and memory costs of many known methods such as hillclimbing, dynamic programming and sampling variable orderings. Secondly, a branch and bound algorithm is presented that integrates parameter and structural constraints with data in a way to guarantee global optimality with respect to the score function. It is an anytime procedure because, if stopped, it provides the best current solution and an estimation about how far it is from the global solution. We show empirically the advantages of the properties and the constraints, and the applicability of the algorithm to large data sets (up to one hundred variables) that cannot be handled by other current methods (limited to around 30 variables). 1.
GADGET SVM: a GossipbAseD subGradiEnT SVM solver
"... Distributed environments such as federated databases, wireless and sensor networks, PeertoPeer (P2P) networks are becoming increasingly popular and wellsuited for machine learning since they can store large ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
(Show Context)
Distributed environments such as federated databases, wireless and sensor networks, PeertoPeer (P2P) networks are becoming increasingly popular and wellsuited for machine learning since they can store large
Generalized Domains for Empirical Evaluations in Reinforcement Learning
"... Many empirical results in reinforcement learning are based on a very small set of environments. These results often represent the best algorithm parameters that were found after an adhoc tuning or fitting process. We argue that presenting tuned scores from a small set of environments leads to metho ..."
Abstract

Cited by 2 (2 self)
 Add to MetaCart
(Show Context)
Many empirical results in reinforcement learning are based on a very small set of environments. These results often represent the best algorithm parameters that were found after an adhoc tuning or fitting process. We argue that presenting tuned scores from a small set of environments leads to method overfitting, wherein results may not generalize to similar environments. To address this problem, we advocate empirical evaluations using generalized domains: parameterized problem generators that explicitly encode variations in the environment to which the learner should be robust. We argue that evaluating across a set of these generated problems offers a more meaningful evaluation of reinforcement learning algorithms. 1.
permission. Fast Support Vector Machine Training and Classification on Graphics Processors
, 2008
"... All rights reserved. ..."
(Show Context)
Efficient Euclidean Projections onto the Intersection of Norm Balls
"... Using sparseinducing norms to learn robust models has received increasing attention from many fields for its attractive properties. Projectionbased methods have been widely applied to learning tasks constrained by such norms. As a key building block of these methods, an efficient operator for Eucl ..."
Abstract
 Add to MetaCart
(Show Context)
Using sparseinducing norms to learn robust models has received increasing attention from many fields for its attractive properties. Projectionbased methods have been widely applied to learning tasks constrained by such norms. As a key building block of these methods, an efficient operator for Euclidean projection onto the intersection of ℓ1 and ℓ1,q norm balls (q = 2 or ∞) is proposed in this paper. We prove that the projection can be reduced to finding the root of an auxiliary function which is piecewise smooth and monotonic. Hence, a bisection algorithm is sufficient to solve the problem. We show that the time complexity of our solution is O(n + g log g) for q = 2 and O(n log n) for q = ∞, where n is the dimensionality of the vector to be projected and g is the number of disjoint groups; we confirm this complexity by experimentation. Empirical study reveals that our method achieves significantly better performance than classical methods in terms of running time and memory usage. We further show that embedded with our efficient projection operator, projectionbased algorithms can solve regression problems with composite norm constraints more efficiently than other methods and give superior accuracy. 1.
Graduate Group ChairpersonCOPYRIGHT
, 2009
"... I would like to acknowledge first and foremost my advisor, Ben Taskar, who represents in my eyes what a perfect advisor should be. As his first Ph.D. student I benefited from a lot of his guidance, his knowledge in machine learning and his inspiration for tackling difficult problems. I am particular ..."
Abstract
 Add to MetaCart
I would like to acknowledge first and foremost my advisor, Ben Taskar, who represents in my eyes what a perfect advisor should be. As his first Ph.D. student I benefited from a lot of his guidance, his knowledge in machine learning and his inspiration for tackling difficult problems. I am particularly grateful to him for helping me select a very interesting thesis topic, and for his constant push for quality, original and ambitious work. Ben was and still is a mentor for me. I would like to thank my thesis committee, Kostas Daniilidis, Fernando Pereira, CJ Taylor and Andrew Zisserman, for their excellent feedback, questions and suggestions which shaped this thesis to its current form. I have found in their work a source of inspiration and knowledge. The work of Andrew on character naming in video using screenplay was particularly influential for me, as was the work of Fernando on Conditional Random Fields. As my committee chair, Kostas gave me invaluable advice to improve my thesis, as well as academic guidance throughout my Ph.D. I would also like to thank my written preliminary exam committee, Ali Jadbabaie, Kostas Daniilidis and Lyle Ungar. I had the chance to work with Eleni Miltsakaki, Michael Shilman, Paul Viola, and especially Jianbo Shi, who taught me a lot about computer vision, how to think hard about a problem and how to ask the right questions before proposing a solution. Jianbo has greatly influenced my research, presentation skills, and approach to problem solving. As a Grasp lab alumni, I have had the privilege to collaborate, work with, or interact with a number of outstanding people. My special thanks go to Ben Sapp, who was an amazing colleague. I would also like to thank Akash Nagle, Chris Jordan, Praveen