In the standard formulation of supervised learning the input is represented as a vector of features. However, in most real-life problems, we also have additional information about each of the features. This information can be represented as a set of properties, referred to as meta-features. For instance, in an image recognition task, where the features are pixels, the meta-features can be the (...

Ontologies represent an important source of prior information which lends itself to the integration into statistical modeling. This paper discusses approaches towards employing ontological knowledge for relational learning. Our analysis is based on the IHRM model that performs relational learning by including latent variables that can be interpreted as cluster variables of the entities in the d...

Prior knowledge has been shown very useful to address many natural language processing tasks. Many approaches have been proposed to formalise a variety of knowledge, however, whether the proposed approach is robust or sensitive to the knowledge supplied to the model has rarely been discussed. In this paper, we propose three regularization terms on top of generalized expectation criteria, and co...

ion Domain layer Inference layer Task layer Strategy layer Figure 2: Degree of abstraction theoretical view When comparing the layers according to degree of abstraction, the above ordering suggests the idea of Fig. 2. However, the design of the strategy layer is, in practice, tightly bound with the problem domain, and can be viewed as more specific than the inference and task layers. Fig. 3 is ...

A new method to pre-segment images by means of a hierarchical description is proposed. This description is obtained from an investigation of the deep structure of a scale space image – the input image and the Gaussian filtered ones simultaneously. We concentrate on scale space critical points – points with vanishing gradient with respect to both spatial and scale direction. We show that these p...

The paper presents a novel method for transfer learning through prior variable sampling. A set of problems defined in the same feature space with similar dependencies of target on features is considered. We suggest a method for learning a decision tree ensemble on each of the problems by prior estimation of variable importance on other problems in the set and using it for regularizing model lea...

This study explores the benefits of integrating knowledge representations in prior art patent retrieval. Key to the introduced approach is the utilization of human judgment available in the form of classifications assigned to patent documents. The paper first outlines in detail how a methodology for the extraction of knowledge from such an hierarchical classification system can be established. ...

We explore methods for incorporating prior knowledge about a problem at hand in Support Vector learning machines. We show that both invariances under group transfonnations and prior knowledge about locality in images can be incorporated by constructing appropriate kernel functions.

The key step in reading is to recognize the word by combining visually acquired letter information with prior knowledge of the possible words, as determined by narrative context and vocabulary. How are these two kinds of information combined to identify a word? The purpose of this project was to determine the effect that vocabulary size has on word identification. Vocabularies of 4, 26 and 1708...

In the note the class of block-oriented dynamic nonlinear systems is considered, in particular, Hammerstein and Wiener systems are investigated. Several algorithms for nonlinear system identification are presented. The algorithms exploit various degrees of prior knowledge from parametric to nonparametric. Eventually, a semiparametric algorithm, which shares advantages of both approaches is anno...