Computationally Efficient Methods for MDL-Optimal Density Estimation and Data Clustering

The Minimum Description Length (MDL) principle is a general, well-founded theoretical formalization of statistical modeling. The most important notion of MDL is the stochastic complexity, which can be interpreted as the shortest description length of a given sample of data relative to a model class. The exact definition of the stochastic complexity has gone through several evolutionary steps. The latest instantation is based on the so-called Normalized Maximum Likelihood (NML) distribution which has been shown to possess several important theoretical properties. However, the applications of this modern version of the MDL have been quite rare because of computational complexity problems, i.e., for discrete data, the definition of NML involves an exponential sum, and in the case of continuous data, a multi-dimensional integral usually infeasible to evaluate or even approximate accurately. In this doctoral dissertation, we present mathematical techniques for computing NML efficiently for some model families involving discrete data. We also show how these techniques can be used to apply MDL in two practical applications: histogram density estimation and clustering of multi-dimensional data.

Clustering Users of Online Content Service

Online content services can greatly benefit from personalisation features that enable delivery of content that is suited to each user's specific interests. This thesis presents a system that applies text analysis and user modeling techniques in an online news service for the purpose of personalisation and user interest analysis. The system creates a detailed thematic profile for each content item and observes user's actions towards content items to learn user's preferences. A handcrafted taxonomy of concepts, or ontology, is used in profile formation to extract relevant concepts from the text. User preference learning is automatic and there is no need for explicit preference settings or ratings from the user. Learned user profiles are segmented into interest groups using clustering techniques with the objective of providing a source of information for the service provider. Some theoretical background for chosen techniques is presented while the main focus is in finding practical solutions to some of the current information needs, which are not optimally served with traditional techniques.

Test of the Clustering Hypothesis in the Helsinki Exchanges

This paper investigates the clustering pattern in the Finnish stock market. Using trading volume and time as factors capturing the clustering pattern in the market, the Keim and Madhavan (1996) and the Engle and Russell (1998) model provide the framework for the analysis. The descriptive and the parametric analysis provide evidences that an important determinant of the famous U-shape pattern in the market is the rate of information arrivals as measured by large trading volumes and durations at the market open and close. Precisely, 1) the larger the trading volume, the greater the impact on prices both in the short and the long run, thus prices will differ across quantities. 2) Large trading volume is a non-linear function of price changes in the long run. 3) Arrival times are positively autocorrelated, indicating a clustering pattern and 4) Information arrivals as approximated by durations are negatively related to trading flow.

Wood and fiber properties of Norway spruce and its suitability for thermomechanical pulping.

In the first part of the study, the selected wood and fiber properties were investigated in terms of their occurrence and variation in wood, as well as their relevance from the perspective of thermomechanical pulping process and related end-products. It was concluded that the most important factors were the fiber dimensions, juvenile wood content, and in some cases, the content of heartwood being associated with extremely dry wood with low permeability in spruce. With respect to the above properties, the following three pulpwood assortments of which pulping potential was assumed to vary were formed: wood from regeneration cuttings, first-thinnings wood, and sawmill chips. In the experimental part of the study the average wood and fiber characteristics and their variation were determined for each raw material group prior to pulping. Subsequently, each assortment - equaling about 1500 m3 roundwood - was pulped separately for a 24 h period, at constant process conditions. The properties of obtained newsgrade thermomechanical pulps were then determined. Thermomechanical pulping (TMP) from sawmill chips had the highest proportion of long fibers, smallest proportion of fines, and had generally the coarsest and longest fibers. TMP from first-thinnings wood was just the opposite, whereas that from regeneration cuttings fell in between the above two extremes. High proportion of dry heartwood in wood originating from regeneration cuttings produced a slightly elevated shives content. However, no differences were found in pulp specific energy consumption. The obtained pulp tear index was clearly best in TMP made from sawmill chips and poorest in pulp from first-thinnings wood, which had generally inferior strength properties. No dramatical differences in any of the strength properties were found between pulp from sawmill residual wood and regeneration cuttings. Pulp optical properties were superior in TMP from first-thinnings. Unexpectedly, no noticeable differences, which could be explained with fiber morphology, were found in sheet density, bulk, air permeance or roughness between the three pulps. The most important wood quality factors in this study were the fiber length, fiber cross-sectional dimensions and percentage juvenile wood. Differences found in the quality of TMP manufactured from the above spruce assortments suggest that they could be segregated and pulped separately to obtain specific product characteristics, i.e., for instance tailor-made end-products, and to minimize unnecessary variation in the raw material quality, and hence, pulp quality.