Pairwise comparison

About: Pairwise comparison is a research topic. Over the lifetime, 6804 publications have been published within this topic receiving 174081 citations.

The appropriateness of analysis of variance and multiple-comparison procedures.

Abstract: Consider the situation where samples have been obtained randomly from each of five populations A, B, C, D and E. The question arises, 'How should the data be analyzed?'. A common answer to this question is 'Do one-way analysis of variance and, if P < .05, make pairwise comparisons using an appropriate multiple-comparison algorithm'. This is the approach presented in virtually all textbooks, subscribed to by most statisticians, and expected (if not required) by most journals. Conversely, the alternative answer, 'Do 10 two-sample t tests', would typically be considered naive and likely to elicit the response, 'You should consult a statistician'. Perhaps most importantly, medical research is likely to fare poorly among reviewers when the statistical analysis is judged to be naive. Thus the purpose of this

Efficient pairwise RNA structure prediction and alignment using sequence alignment constraints

Abstract: We are interested in the problem of predicting secondary structure for small sets of homologous RNAs, by incorporating limited comparative sequence information into an RNA folding model. The Sankoff algorithm for simultaneous RNA folding and alignment is a basis for approaches to this problem. There are two open problems in applying a Sankoff algorithm: development of a good unified scoring system for alignment and folding and development of practical heuristics for dealing with the computational complexity of the algorithm. We use probabilistic models (pair stochastic context-free grammars, pairSCFGs) as a unifying framework for scoring pairwise alignment and folding. A constrained version of the pairSCFG structural alignment algorithm was developed which assumes knowledge of a few confidently aligned positions (pins). These pins are selected based on the posterior probabilities of a probabilistic pairwise sequence alignment. Pairwise RNA structural alignment improves on structure prediction accuracy relative to single sequence folding. Constraining on alignment is a straightforward method of reducing the runtime and memory requirements of the algorithm. Five practical implementations of the pairwise Sankoff algorithm – this work (Consan), David Mathews' Dynalign, Ian Holmes' Stemloc, Ivo Hofacker's PMcomp, and Jan Gorodkin's FOLDALIGN – have comparable overall performance with different strengths and weaknesses.

Jointly Identifying Temporal Relations with Markov Logic

Abstract: Recent work on temporal relation identification has focused on three types of relations between events: temporal relations between an event and a time expression, between a pair of events and between an event and the document creation time. These types of relations have mostly been identified in isolation by event pairwise comparison. However, this approach neglects logical constraints between temporal relations of different types that we believe to be helpful. We therefore propose a Markov Logic model that jointly identifies relations of all three relation types simultaneously. By evaluating our model on the TempEval data we show that this approach leads to about 2% higher accuracy for all three types of relations ---and to the best results for the task when compared to those of other machine learning based systems.

Balancing exploration and exploitation in listwise and pairwise online learning to rank for information retrieval

Abstract: As retrieval systems become more complex, learning to rank approaches are being developed to automatically tune their parameters. Using online learning to rank, retrieval systems can learn directly from implicit feedback inferred from user interactions. In such an online setting, algorithms must obtain feedback for effective learning while simultaneously utilizing what has already been learned to produce high quality results. We formulate this challenge as an exploration---exploitation dilemma and propose two methods for addressing it. By adding mechanisms for balancing exploration and exploitation during learning, each method extends a state-of-the-art learning to rank method, one based on listwise learning and the other on pairwise learning. Using a recently developed simulation framework that allows assessment of online performance, we empirically evaluate both methods. Our results show that balancing exploration and exploitation can substantially and significantly improve the online retrieval performance of both listwise and pairwise approaches. In addition, the results demonstrate that such a balance affects the two approaches in different ways, especially when user feedback is noisy, yielding new insights relevant to making online learning to rank effective in practice.