Relevance feedback is often a critical component when designing image databases. With these databases it is difficult to specify queries directly and explicitly. Relevance feedback interactively determinines a user’s desired output or query concept by asking the user whether certain proposed images are relevant or not. For a relevance feedback algorithm to be effective, it must grasp a user’s query concept accurately and quickly, while also only asking the user to label a small number of images. We propose the use of a support vector machine active learning algorithm for conducting effective relevance feedback for image retrieval. The algorithm selects the most informative images to query a user and quickly learns a boundary that separates the images that satisfy the user’s query concept from the rest of the dataset. Experimental results show that our algorithm achieves significantly higher search accuracy than traditional query refinement schemes after just three to four rounds of relevance feedback.

Recently there has been significant interest in supervised learning algorithms that combine labeled and unlabeled data for text learning tasks. The co-training setting [1] applies to datasets that have a natural separation of their features into two disjoint sets. We demonstrate that when learning from labeled and unlabeled data, algorithms explicitly leveraging a natural independent split of the features outperform algorithms that do not. When a natural split does not exist, co-training algorithms that manufacture a feature split may out-perform algorithms not using a split. These results help explain why co-training algorithms are both discriminative in nature and robust to the assumptions of their embedded classifiers. Categories and Subject Descriptors I.2.6 [Artificial Intelligence]: Learning; H.3.3 [Information Storage and Retrieval]: Information Search and Retrieval--- Information Filtering Keywords co-training, expectation-maximization, learning with labeled and unlabeled...

There has been growing interest in practice in using unlabeled data together with labeled data in machine learning, and a number of di#erent approaches have been developed. However, the assumptions these methods are based on are often quite distinct and not captured by standard theoretical models. In this paper we describe a PAC-style framework that can be used to model many of these assumptions, and analyze sample-complexity issues in this setting: that is, how much of each type of data one should expect to need in order to learn well, and what are the basic quantities that these numbers depend on. Our model can be viewed as an extension of the standard PAC model, where in addition to a concept class C, one also proposes a type of compatibility that one believes the target concept should have with the underlying distribution.

Supervised learning techniques for text classification often require a large number of labeled examples to learn accurately. One way to reduce the amount of labeled data required is to develop algorithms that can learn effectively from a small number of labeled examples augmented with a large number of unlabeled examples. Current text learning techniques for combining labeled and unlabeled, such as EM and Co-Training, are mostly applicable for classification tasks with a small number of classes and do not scale up well for large multiclass problems. In this paper, wedevelop a framework to incorporate unlabeled data in the Error-Correcting Output Coding (ECOC) setup by first decomposing multiclass problems into multiple binary problems and then using Co-Training to learn the individual binary classification problems.

Some classes of learning problems have been well-posed and investigated, especially the ones of classi cation and regression. However, in practice we are often confronted with modi ed learning tasks that deviate from these standard scenarios. In other words, given an application problem, the assumptions made when treating it as a standard learning task are often not appropriate.

We address the problem of using partially labelled data, eg large collections were only little data is annotated, for extracting biological entities. Our approach relies on a combination of probabilistic models, whichwe use to model the generation of entities and their context, and kernel machines, which implementpowerful categorisers based on a similarity measure and some labelled data. This combination takes the form of the so-called Fisher kernels which implement a similarity based on an underlying probabilistic model. Suchkernels are compared with transductive inference, an alternative approachto combining labelled and unlabelled data, again coupled with Support Vector Machines. Experiments are performed on a database of abstracts extracted from Medline.

In this editorial we brie ¯ y discuss interaction between two important areas of arti ® cial intelligence: computer vision (CV) and machine learning (ML). Although the two ® elds have a long-standing tradition and can be considered technologically mature, past research in applying ML techniques to CV problems has been limited. After a short introduction in the ® elds of computer vision and machine learning, we highlight some important issues in the intersection of the two areas and sketch both current achievements and future research directions. Our goal is to help the reader put the six contributions of this special issue into the proper context. In recent years, arti ® cial intelligence has been viewed as the study and construction of rational agents, where the term agent simply denotes something that perceives its environment through sensors and acts upon the environment through e � ectors (Russel & Norvig, 1995). Many sensory modalities have been made available to arti ® cial agents, the most useful of which is certainly vision. In vision the sensory stimulus processed is the light scattered from objects in a scene and projected in two-dimensional images. In computer vision (CV), an image is represented as an array of pixels, which is a twodimensional function f …x;y † that returns the intensity recorded by a light sensor at all points (x;y) of an image plane. The task of a computer vision system is to understand the scene that an image depicts. The development of computer vision systems requires the solution to several problems, since the information to be processed, which is images, is