An experimental study on two decision issues for wrapper feature selection (FS) with multilayer perceptrons and the sequential backward selection (SBS) procedure is presented. The decision issues studied are the stopping criterion and the network retraining before computing the saliency. Experimental results indicate that the increase in the computational cost associated with retraining the network with every feature temporarily removed before computing the saliency is rewarded with a significant performance improvement. Despite being quite intuitive, this idea has been hardly used in practice. A somehow nonintuitive conclusion can be drawn by looking at the stopping criterion, suggesting that forcing overtraining may be as useful as early stopping. A significant improvement in the overall results with respect to learning with the whole set of variables is observed.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1109/TNN.2007.909535 | DOI Listing |
Curr Med Chem
January 2025
Shree S K Patel College of Pharmaceutical Education and Research, Ganpat University, Mahesana, Gujarat, 384012, India.
Therapeutic hurdles persist in the fight against lung cancer, although it is a leading cause of cancer-related deaths worldwide. Results are still not up to par, even with the best efforts of conventional medicine, thus new avenues of investigation are required. Examining how immunotherapy, precision medicine, and AI are being used to manage lung cancer, this review shows how these tools can change the game for patients and increase their chances of survival.
View Article and Find Full Text PDFActa Med Philipp
December 2024
Department of Physics, School of Science and Engineering, Ateneo de Manila University, Quezon City, Philippines.
Background And Objective: The adoption of electronic medical records (EMRs) in the Philippines has been initiated and adjusted since the last decade through the Philippine eHealth Agenda framework. EMRs are known to improve clinical management and have been widely adopted in advanced economies. However, empirical research on EMR implementation remains limited.
View Article and Find Full Text PDFJACC Adv
December 2024
Department of Cardiology, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China.
Background: Plant-based dietary patterns are becoming increasingly popular due to environmental and health impacts, yet there are few studies exploring the relationship between plant-based dietary patterns and venous thromboembolism (VTE) in different genetic backgrounds.
Objectives: The aim of this study was to investigate how plant-based dietary pattern and genetic susceptibility independently or jointly affect VTE and its subtypes of pulmonary embolism and deep vein thrombosis.
Methods: A total of 183,510 participants who were White British ethnicity background and free of VTE at baseline in the UK Biobank were recruited, in consideration that the selection of genetic variants for VTE was based on results of White European individuals.
ACS Pharmacol Transl Sci
January 2025
Institute of Pharmaceutical and Medicinal Chemistry, University of Münster, 48149 Münster, Germany.
This study presents a novel series of -acylated 1,2,4-triazol-5-amines and 1-pyrazol-5-amines, featuring a pyrazin-2-yl moiety, developed as covalent inhibitors of thrombin. These compounds demonstrated potent inhibitory activity, with derivatives and achieving IC values as low as 0.7 and 0.
View Article and Find Full Text PDFJ Spine Surg
December 2024
Department of Neurosurgery, The Gemelli University Hospital, Rome, Italy.
Background: Aneurysmal bone cysts (ABCs) are benign, blood-filled neoplasms causing bone destruction, often requiring resection. However, challenges arise, especially at the cranio-cervical junction, where proximity to critical structures limits removal. Non-surgical options include selective arterial embolization (SAE) as main treatment, while Denosumab and centrifugated bone marrow emerge as experimental alternatives.
View Article and Find Full Text PDFEnter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!