Helsinki Institute for Information Technology

About: Helsinki Institute for Information Technology is a facility organization based out in Espoo, Finland. It is known for research contribution in the topics: Population & Bayesian network. The organization has 630 authors who have published 1962 publications receiving 63426 citations.

Making the ordinary visible in microblogs

Abstract: Microblogging is a "Mobile Web 2.0" service category that enables brief blog-like postings from mobile terminals and PCs to the World Wide Web. To shed light on microblogging as a communication genre, we report on multiple analyses of data from the first 10 months of a service called Jaiku. The main finding is that microblogging centers on selective, I-centered disclosure of current activities and experiences, making daily experiences visible for others. The high frequency of brief and mundane status updates, like "working," may be a second-order effect resulting from posting becoming a routine executed to keep the audience interested. The results highlight the importance of reciprocal activity and feedback in users' motivation to invest in this activity.

Identifying Meaningful Places: The Non-parametric Way

Abstract: Gathering and analyzing location data is an important part of many ubiquitous computing applications. The most common way to represent location information is to use numerical coordinates, e.g., latitudes and longitudes. A problem with this approach is that numerical coordinates are usually meaningless to a user and they contrast with the way humans refer to locations in daily communication. Instead of using coordinates, humans tend to use descriptive statements about their location; for example, "I'm home" or "I'm at Starbucks." Locations, to which a user can attach meaningful and descriptive semantics, are often called places. In this paper we focus on the automatic extraction of places from discontinuous GPS measurements. We describe and evaluate a non-parametric Bayesian approach for identifying places from this kind of data. The main novelty of our approach is that the algorithm is fully automated and does not require any parameter tuning. Another novel aspect of our algorithm is that it can accurately identify places without temporal information. We evaluate our approach using data that has been gathered from different users and different geographic areas. The traces that we use exhibit different characteristics and contain data from daily life as well as from traveling abroad. We also compare our algorithm against the popular k-means algorithm. The results indicate that our method can accurately identify meaningful places from a variety of location traces and that the algorithm is robust against noise.

Fast distributed approximation algorithms for vertex cover and set cover in anonymous networks

Abstract: We present a distributed algorithm that finds a maximal edge packing in O(Δ + log* W) synchronous communication rounds in a weighted graph, independent of the number of nodes in the network; here Δ is the maximum degree of the graph and W is the maximum weight. As a direct application, we have a distributed 2-approximation algorithm for minimum-weight vertex cover, with the same running time. We also show how to find an $f$-approximation of minimum-weight set cover in O(f2k2 + fk log* W) rounds; here k is the maximum size of a subset in the set cover instance, f is the maximum frequency of an element, and W is the maximum weight of a subset. The algorithms are deterministic, and they can be applied in anonymous networks.

On the Hyperprior Choice for the Global Shrinkage Parameter in the Horseshoe Prior

Abstract: The horseshoe prior has proven to be a noteworthy alternative for sparse Bayesian estimation, but as shown in this paper, the results can be sensitive to the prior choice for the global shrinkage hyperparameter. We argue that the previous default choices are dubious due to their tendency to favor solutions with more unshrunk coefficients than we typically expect a priori. This can lead to bad results if this parameter is not strongly identified by data. We derive the relationship between the global parameter and the effective number of nonzeros in the coefficient vector, and show an easy and intuitive way of setting up the prior for the global parameter based on our prior beliefs about the number of nonzero coefficients in the model. The results on real world data show that one can benefit greatly -- in terms of improved parameter estimates, prediction accuracy, and reduced computation time -- from transforming even a crude guess for the number of nonzero coefficients into the prior for the global parameter using our framework.

A System-Level Model for Runtime Power Estimation on Mobile Devices

Abstract: The growing popularity of mobile internet services, characterized by heavy network transmission, intensive computation and an always-on display, poses a great challenge to the battery lifetime of mobile devices. To manage the power consumption in an efficient way, it is essential to understand how the power is consumed at the system level and to be able to estimate the power consumption during runtime. Although the power modeling of each hardware component has been studied separately, there is no general solution at present of combining them into a system-level power model. In this paper we present a methodology for building a system-level power model without power measurement at the component level. We develop a linear regression model with nonnegative coefficients, which describes the aggregate power consumption of the processors, the wireless network interface and the display. Based on statistics and expert knowledge, we select three hardware performance counters, three network transmission parameters and one display parameter as regression variables. The power estimation, based on our model, exhibits 2.62 percent median error on real mobile internet services.