sminchisescu.ins.uni-bonn.de In visual recognition, the images are frequently modeled as unordered collections of local features (bags). We show that bag-of-words representations commonly used in conjunction with linear classifiers can be viewed as special match kernels, which count 1 if two local features fall into the same regions partitioned by visual words and 0 otherwise. Despite its simplicity, this quantization is too coarse, motivating research into the design of match kernels that more accurately measure the similarity between local features. However, it is impractical to use such kernels for large datasets due to their significant computational cost. To address this problem, we propose efficient match kernels (EMK) that map local features to a low dimensional feature space and average the resulting vectors to form a setlevel feature. The local feature maps are learned so their inner products preserve, to the best possible, the values of the specified kernel function. Classifiers based on EMK are linear both in the number of images and in the number of local features. We demonstrate that EMK are extremely efficient and achieve the current state of the art in three difficult computer vision datasets: Scene-15, Caltech-101 and Caltech-256. 1

The Fisher kernel (FK) is a generic framework which combines the benefits of generative and discriminative approaches. In the context of image classification the FK was shown to extend the popular bag-of-visual-words (BOV) by going beyond count statistics. However, in practice, this enriched representation has not yet shown its superiority over the BOV. In the first part we show that with several well-motivated modifications over the original framework we can boost the accuracy of the FK. On PASCAL VOC 2007 we increase the Average Precision (AP) from 47.9 % to 58.3%. Similarly, we demonstrate state-of-the-art accuracy on CalTech 256. A major advantage is that these results are obtained using only SIFT descriptors and costless linear classifiers. Equipped with this representation, we can now explore image classification on a larger scale. In the second part, as an application, we compare two abundant resources of labeled images to learn classifiers: ImageNet and Flickr groups. In an evaluation involving hundreds of thousands of training images we show that classifiers learned on Flickr groups perform surprisingly well (although they were not intended for this purpose) and that they can complement classifiers learned on more carefully annotated datasets.

This paper explores an important and relatively unstudied quality measure of a sponsored search advertisement: bounce rate. The bounce rate of an ad can be informally defined as the fraction of users who click on the ad but almost immediately move on to other tasks. A high bounce rate can lead to poor advertiser return on investment, and suggests search engine users may be having a poor experience following the click. In this paper, we first provide quantitative analysis showing that bounce rate is an effective measure of user satisfaction. We then address the question, can we predict bounce rate by analyzing the features of the advertisement? An affirmative answer would allow advertisers and search engines to predict the effectiveness and quality of advertisements before they are shown. We propose solutions to this problem involving large-scale learning methods that leverage features drawn from ad creatives in addition

A wide variety of machine learning problems can be described as minimizing a regularized risk functional, with different algorithms using different notions of risk and different regularizers. Examples include linear Support Vector Machines (SVMs), Gaussian Processes, Logistic Regression, Conditional Random Fields (CRFs), and Lasso amongst others. This paper describes the theory and implementation of a scalable and modular convex solver which solves all these estimation problems. It can be parallelized on a cluster of workstations, allows for data-locality, and can deal with regularizers such as L1 and L2 penalties. In addition to the unified framework we present tight convergence bounds, which show that our algorithm converges in O(1/ɛ) steps to ɛ precision for general convex problems and in O(log(1/ɛ)) steps for continuously differentiable problems. We demonstrate the performance of our general purpose solver on a variety of publicly available datasets.

Online learning algorithms have impressive convergence properties when it comes to risk minimization and convex games on very large problems. However, they are inherently sequential in their design which prevents them from taking advantage of modern multi-core architectures. In this paper we prove that online learning with delayed updates converges well, thereby facilitating parallel online learning. 1

Structured prediction tasks pose a fundamental trade-off between the need for model complexity to increase predictive power and the limited computational resources for inference in the exponentially-sized output spaces such models require. We formulate and develop structured prediction cascades: a sequence of increasingly complex models that progressively filter the space of possible outputs. We represent an exponentially large set of filtered outputs using max marginals and propose a novel convex loss function that balances filtering error with filtering efficiency. We provide generalization bounds for these loss functions and evaluate our approach on handwriting recognition and part-of-speech tagging. We find that the learned cascades are capable of reducing the complexity of inference by up to five orders of magnitude, enabling the use of models which incorporate higher order features and yield higher accuracy. 1

We describe and analyze two stochastic methods for ℓ1 regularized loss minimization problems, such as the Lasso. The first method updates the weight of a single feature at each iteration while the second method updates the entire weight vector but only uses a single training example at each iteration. In both methods, the choice of feature/example is uniformly at random. Our theoretical runtime analysis suggests that the stochastic methods should outperform state-of-the-art deterministic approaches, including their deterministic counterparts, when the size of the problem is large. We demonstrate the advantage of stochastic methods by experimenting with synthetic and natural data sets. 1.

Many problems arising in machine learning can be cast as a convex optimization problem, in which a sum of a loss term and a regularization term is minimized. For example, in Support Vector Machines the loss term is the average hinge-loss of a vector over a training set of examples and the regularization term is the squared Euclidean norm of this vector. In this paper we study an algorithmic framework for strongly convex repeated games and apply it for solving regularized loss minimization problems. In a convex repeated game, a predictor chooses a sequence of vectors from a convex set. After each vector is chosen, the opponent responds with a convex loss function and the predictor pays for applying the loss function to the vector she chose. The regret of the predictor is the difference between her cumulative loss and the minimal cumulative loss achievable by a fixed vector, even one that is chosen in hindsight. In strongly convex repeated games, the opponent is forced to choose loss functions that are strongly convex. We describe a family of prediction algorithms for strongly convex repeated games that attain logarithmic regret. 1

While large-scale taxonomies – especially for web pages – have been in existence for some time, approaches to automatically classify documents into these taxonomies have met with limited success compared to the more general progress made in text classification. We argue that this stems from three causes: increasing sparsity of training data at deeper nodes in the taxonomy, error propagation where a mistake made high in the hierarchy cannot be recovered, and increasingly complex decision surfaces in higher nodes in the hierarchy. While prior research has focused on the first problem, we introduce methods that target the latter two problems – first by biasing the training distribution to reduce error propagation and second by propagating up “first-guess ” expert information in a bottom-up manner before making a refined top down choice. Finally, we present an empirical study demonstrating that the suggested changes lead to 10-30 % improvements in F1 scores versus an accepted competitive baseline, hierarchical SVMs.

This paper examines the generalization properties of online convex programming algorithms when the loss function is Lipschitz and strongly convex. Our main result is a sharp bound, that holds with high probability, on the excess risk of the output of an online algorithm in terms of the average regret. This allows one to use recent algorithms with logarithmic cumulative regret guarantees to achieve fast convergence rates for the excess risk with high probability. As a corollary, we characterize the convergence rate of PEGASOS (with high probability), a recently proposed method for solving the SVM optimization problem. 1