A5059446980- Data Visualization and Analytics
- Neural Networks and Applications
- Face and Expression Recognition
- Topic Modeling
- Information Retrieval and Search Behavior
- Data Management and Algorithms
- Bayesian Methods and Mixture Models
- Natural Language Processing Techniques
- Complex Network Analysis Techniques
- Machine Learning and Data Classification
- Advanced Text Analysis Techniques
- Digital Games and Media
- Diabetes and associated disorders
- Image Retrieval and Classification Techniques
- Personal Information Management and User Behavior
- Computational and Text Analysis Methods
- Domain Adaptation and Few-Shot Learning
- Topological and Geometric Data Analysis
- Emergency and Acute Care Studies
- Video Analysis and Summarization
- Statistical Methods in Epidemiology
- Advanced Statistical Methods and Models
- Geographic Information Systems Studies
- Linguistic Variation and Morphology
- Mobile Crowdsensing and Crowdsourcing
Tampere University
2015-2024
Aalto University
2010-2020
Helsinki Institute for Information Technology
2007-2017
University of Helsinki
2001-2005
University of Oulu
2002-2003
Dimensionality Reduction (DR) is a core building block in visualizing multidimensional data. For DR techniques to be useful exploratory data analysis, they need adapted human needs and domain-specific problems, ideally, interactively, on-the-fly. Many visual analytics systems have already demonstrated the benefits of tightly integrating with interactive visualizations. Nevertheless, general, structured understanding this integration missing. To address this, we systematically studied...
Genes with similar transcriptional activation kinetics can display very different temporal mRNA profiles because of differences in transcription time, degradation rate, and RNA-processing kinetics. Recent studies have shown that a splicing-associated RNA production delay be significant. To investigate this issue more generally, it is useful to develop methods applicable genome-wide datasets. We introduce joint model accumulation used for inference delay, rate given data from high-throughput...
We introduce a method for deriving metric, locally based on the Fisher information matrix, into data space. A self-organizing map (SOM) is computed in new metric to explore financial statements of enterprises. The measures local distances terms changes distribution an auxiliary random variable that reflects what important data. In this paper indicates bankruptcy within next few years. conditional density first estimated, and change estimate resulting from displacements primary space measured...
We introduce interactive intent modeling, where the user directs exploratory search by providing feedback for estimates of intents. The estimated intents are visualized interaction on an Intent Radar, a novel visual interface that organizes onto radial layout relevant close to center visualization and similar have angles. can give intents, from which system learns visualizes improved estimates. systematically evaluated effect modeling in mixed-method task-based information seeking setting...
Exploratory search requires the system to assist user in comprehending information space and expressing evolving intents for iterative exploration retrieval of information. We introduce interactive intent modeling, a technique that models user’s visualizes them as keywords interaction. The can provide feedback on keywords, from which learns an improved estimate retrieves report experiments comparing variants implementing modeling control system. Data comprising logs, interaction essay...
Importance Magnetic Resonance Imaging (MRI) coupled with Prostate Imaging-Reporting and Data System (PI-RADS) provides a standardized scoring system for assessing clinical significance of prostate cancer (PCa). However, the association between PI-RADS scores key end-points remains underexplored due to limited follow-up data. Objective To evaluate cancer-specific mortality (PCSM), overall survival (OS), metastasis-free (MFS), biochemical recurrence (BCR) across multiple cohorts. Design,...
Dimensionality reduction is one of the basic operations in toolbox data analysts and designers machine learning pattern recognition systems. Given a large set measured variables but few observations, an obvious idea to reduce degrees freedom measurements by rep resenting them with smaller more "condensed" variables. Another reason for reducing dimensionality computational load further processing. A third visualization.
Abstract Background and objective Emergency Department (ED) overcrowding is a chronic international issue that associated with adverse treatment outcomes. Accurate forecasts of future service demand would enable intelligent resource allocation could alleviate the problem. There has been continued academic interest in ED forecasting but number used explanatory variables low, limited mainly to calendar weather variables. In this study we investigate whether predictive accuracy next day...
We introduce topic-relevance map, an interactive search result visualization that assists rapid information comprehension across a large ranked set of results. The map visualizes topical overview the space as keywords with respect to two essential retrieval measures: relevance and similarity. Non-linear dimensionality reduction is used embed high-dimensional keyword representations data into angles on radial layout. Relevance estimated by ranking method visualized radiuses As result, similar...
The increasing use of artificial intelligence (AI) technologies across application domains has prompted our society to pay closer attention AI's trustworthiness, fairness, interpretability, and accountability. In order foster trust in AI, it is important consider the potential interactive visualization, how such visualizations help build AI systems. This manifesto discusses relevance makes following four claims: i) not a technical problem, ii) dynamic, iii) visualization cannot address all...
A simple probabilistic model is introduced to generalize classical linear discriminant analysis (LDA) in finding components that are informative of or relevant for data classes. The maximize the predictability class distribution which asymptotically equivalent 1) maximizing mutual information with classes, and 2) principal so-called learning Fisher metrics. metric measures only distances is, cause changes distribution. have applications exploration, visualization, dimensionality reduction....
We present a conceptual framework for the development of visual interactive techniques to formalize and externalize trust in machine learning (ML) workflows. Currently, ML applications is an implicit process that takes place user's mind. As such, there no method feedback or communication can be acted upon. Our will instrumental developing visualization approaches help users efficiently effectively build communicate ways fit each stages. formulate several research questions directions...
Human endogenous retroviruses (HERVs) are surviving traces of ancient retrovirus infections and now reside within the human DNA. Recently HERV expression has been detected in both normal tissues diseased patients. However, activities (expression levels) individual sequences mostly unknown. We introduce a generative mixture model, based on Hidden Markov Models, for estimating from EST (expressed sequence tag) databases. use model to estimate relative 181 HERVs. also empirically justify faster...
In difficult information seeking tasks, the majority of top-ranked documents for an initial query may be non-relevant, and negative relevance feedback then help find relevant documents. Traditional has been studied on document results; we introduce a system interface in novel exploratory search setting, where continuous-valued is directly given to keyword features inferred probabilistic user intent model. The introduced allows both positive interactive visual interface, by letting manipulate...
Abstract Computational recognition of narratives, if successful, would find innumerable applications with large digitized datasets. Systematic identification narratives in the text flow could significantly contribute to such pivotal questions as where, when, and how are employed. This paper discusses an approach extract from two datasets, Finnish parliamentary records (1980–2021) oral history interviews former MPs (1988–2018). Our study was based on iterative approach, proceeding original...