PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 15787152

Optimizing the kernel in the empirical feature space.

Huilin Xiong·M N S Swamy·M Omair Ahmad
View original source ↗Export metadata ↓

Abstract

In this paper, we present a method of kernel optimization by maximizing a measure of class separability in the empirical feature space, an Euclidean space in which the training data are embedded in such a way that the geometrical structure of the data in the feature space is preserved. Employing a data-dependent kernel, we derive an effective kernel optimization algorithm that maximizes the class separability of the data in the empirical feature space. It is shown that there exists a close relationship between the class separability measure introduced here and the alignment measure defined recently by Cristianini. Extensive simulations are carried out which show that the optimized kernel is more adaptive to the input data, and leads to a substantial, sometimes significant, improvement in the performance of various data classification algorithms.

Explore related subjects

Empirical ResearchNeural Networks, ComputerJournal ArticleResearch Support, Non-U.S. Gov't

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Huilin Xiong, M N S Swamy, M Omair Ahmad. 2005. Optimizing the kernel in the empirical feature space.. https://doi.org/10.1109/tnn.2004.841784

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2006

Concept empiricism: a methodological critique.

Thanks to Barsalou's, Damasio's, Glenberg's, Prinz' and others' work, neo-empiricism is gaining a deserved recognition in the psychology and philosophy of concepts. I argue, however, that neo-empiricists have underestimated the difficulty of providing evidence against the amodal approach to concepts and higher cognition. I highlight three key problems: the difficulty of sorting out amodal predictions from neo-empiricist predictions, the difficulty of finding experimental tasks that are not best solved by imagery and the difficulty of generalizing findings concerning a given cognitive process in a given context to other cognitive processes or other contexts. Finally, solutions to these three problems are considered.

Empirical Research↗
PubMed2006

Surgery for epilepsy: assessing evidence from observational studies.

The effectiveness of epilepsy surgery is mostly based on the results of observational studies because an experimental design in surgical patients has several ethical implications. However, observational studies suffer from methodological drawbacks, which prevent meaningful conclusions and represent a serious limitation when data from different reports are pooled for systematic reviews and meta-analyses. These include the retrospective design, the use of referral populations, the small sample size, the enrolment of patients at differing inception points, the use of differing inclusion criteria, differing definitions of prognostic predictors and outcome measures, the unmasked assessment of outcome, the short follow-up, the changing technologies and surgical procedures, and the inadequate statistical methods. With these limitations in mind, a systematic review was performed of epilepsy surgery studies aiming at identifying the positive and negative prognostic predictors of surgical outcome. In this review, extent of surgical resection, abnormal MRI, mesial temporal sclerosis, febrile seizures, EEG/MRI concordance, and tumor were in decreasing order the principal indicators of the success of surgery of epilepsy. By contrast, intracranial monitoring and post-operative discharges tended to predict an unfavorable prognosis. Although the heterogeneity of the study results was fairly low, the results may be affected by pooling of data from heterogeneous reports (different patient series assessed with differing methods) or unexamined or unknown confounders. These data must be considered preliminary and cannot replace well-conducted prognostic studies (representative study populations, well-defined inception cohorts, satisfactory and complete follow-up, prospective design, and standard definition of prognostic factors).

Empirical Research↗
PubMed2006

A systematic review of empirical research on self-reported racism and health.

This paper reviews 138 empirical quantitative population-based studies of self-reported racism and health. These studies show an association between self-reported racism and ill health for oppressed racial groups after adjustment for a range of confounders. The strongest and most consistent findings are for negative mental health outcomes and health-related behaviours, with weaker associations existing for positive mental health outcomes, self-assessed health status, and physical health outcomes. Most studies in this emerging field have been published in the past 5 years and have been limited by a dearth of cohort studies, a lack of psychometrically validated exposure instruments, poor conceptualization and definition of racism, conflation of racism with stress, and debate about the aetiologically relevant period for self-reported racism. Future research should examine the psychometric validity of racism instruments and include these instruments, along with objectively measured health outcomes, in existing large-scale survey vehicles as well as longitudinal studies and studies involving children. There is also a need to gain a better understanding of the perception, attribution, and reporting of racism, to investigate the pathways via which self-reported racism affects health, the interplay between mental and physical health outcomes, and exposure to intra-racial, internalized, and systemic racism. Ensuring the quality of studies in this field will allow future research to reveal the complex role that racism plays as a determinant of population health.

Empirical Research↗