Dr. Carlos HUERTAS (Universidad Autonoma de Baja California, Mexico), "18 Temmuz 2016 Pazartesi günü saat 14:00"da Yüksek Lisans dersanesinde bir seminer verecektir.
Seminer başlığı ve özeti aşağıda verilmiştir. Bölümümüz öğretim üyeleri, lisans ve lisans üstü öğrenciler katılabilirler.


Title: Heat Map Based Feature Selection: Dimensionality Reduction Technique Applied in High Dimensional Data

Presenter: Dr. Carlos HUERTAS 

Abstract: Public health is a critical issue, therefore we can find a great research interest to find faster and more accurate methods to detect diseases. Technologies such as mass spectrometry (MS) have improved the way detections are made, however, since MS by nature comes with a very high features-to-samples ratio, several problems arise to find reliable biomarkers to help machine learning techniques automate detection. Nowadays we have several techniques for feature selection that help in the seek for biomarker discovery, nonetheless the selection can be biased towards a particular classifier instead of a more stable selection that could be applied to new data. In this work we present an algorithm called heat map based feature selection (HmbFS), that uses a built in compression mechanism based on color quantization to optimize the selection and help to improve current techniques by assemble of multiple selections.

Eklenme Tarihi: 14.07.2016 22:59:11
İstanbul Üniversitesi Mühendislik Fakültesi Bilgisayar Mühendisliği Bölümü
İstanbul Üniversitesi Mühendislik Fakültesi Avcılar Kampüsü, 34320 Avcılar - İSTANBUL
Tel: +90 212 473 70 70 Faks: +90 212 473 71 80