Multi-view data refers to a setting where features are divided into feature sets, for example because they correspond to different sources. Stacked penalized logistic regression (StaPLR) is a... Show moreMulti-view data refers to a setting where features are divided into feature sets, for example because they correspond to different sources. Stacked penalized logistic regression (StaPLR) is a recently introduced method that can be used for classification and automatically selecting the views that are most important for prediction. We introduce an extension of this method to a setting where the data has a hierarchical multi-view structure. We also introduce a new view importance measure for StaPLR, which allows us to compare the importance of views at any level of the hierarchy. We apply our extended StaPLR algorithm to Alzheimer's disease classification where different MRI measures have been calculated from three scan types: structural MRI, diffusion-weighted MRI, and resting-state fMRI. StaPLR can identify which scan types and which derived MRI measures are most important for classification, and it outperforms elastic net regression in classification performance. Show less
Wiggers, G.; Verberne, S.; Zwenne, G.J.; Loon, W.S. van 2022
This paper addresses relevance in legal information retrieval (IR). We investigate whether the conceptual framework of relevance in legal IR, as described by Van Opijnen (2017), can be confirmed in... Show moreThis paper addresses relevance in legal information retrieval (IR). We investigate whether the conceptual framework of relevance in legal IR, as described by Van Opijnen (2017), can be confirmed in practice. The research is conducted with a user questionnaire in which users of a legal IR system had to choose which of two results they would like to see ranked higher for a query and were asked to provide a reasoning for their choice. To avoid questions with an obvious answer and extract as much information as possible about the reasoning process, the search results were chosen to differ on relevance factors from the literature, where one result scores high on one factor, and the other on another factor. The questionnaire had eleven pairs of search results. A total of 43 legal professionals participated: 14 legal information specialists, 6 legal scholars and 23 legal practitioners.The results confirms the existence of domain relevance as described in the theoretical framework by Van Opijnen (2017). Based on the factors mentioned by the respondents, we can conclude that document type, recency, level of depth, legal hierarchy, authority, usability and whether a document is annotated are factors of domain relevance that are largely independent of the task context.We also investigated whether different sub-groups of users of legal IR systems (legal information specialists who are searching for others, legal scholars, and legal practitioners) differ in terms of the factors they consider in judging the relevance of legal documents outside of a task context. Using a PERMANOVA we found no significant difference in the factors reported by these groups. At this moment there is no reason to treat these sub-groups differently in legal IR systems. Show less
Loon, W.S. van; Fokkema, M.; Szabo, B.T., Rooij, M.J. de 2018
In biomedical research many different types of patient data can be collected, including various types of omics data and medical imaging modalities. Applying multi-view learning to these different... Show moreIn biomedical research many different types of patient data can be collected, including various types of omics data and medical imaging modalities. Applying multi-view learning to these different sources of information can increase the accuracy of medical classification models compared with single-view procedures. However, the collection of biomedical data can be expensive and taxing on patients, so that superfluous data collection should be avoided. It is therefore necessary to develop multi-view learning methods which can accurately identify the views most important for prediction. In recent years, several biomedical studies have used an approach known as multi-view stacking (MVS), where a model is trained on each view separately and the resulting predictions are combined through stacking. In these studies, MVS has been shown to increase classification accuracy. However, the MVS framework can also be used for selecting a subset of important views. To study the view selection potential of MVS, we develop a special case called stacked penalized log istic regression (StaPLR). Compared with existing view-selection methods, StaPLR can make use of faster optimization algorithms and is easily parallelized. We show that nonnegativity constraints on the parameters of the function which combines the views are important for preventing unimportant views from entering the model. We investigate the performance of StaPLR through simulations, and consider two real data examples. We compare the performance of StaPLR with an existing view selection method called the group lasso and observe that, in terms of view selection, StaPLR has a consistently lower false positive rate. Show less
