BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260514T074826EDT-56368JBr9T@132.216.98.100
DTSTAMP:20260514T114826Z
DESCRIPTION:Visualizing emergent patterns for exploratory data analysis\n\n
 High-throughput data collection is becoming increasingly common\, and ofte
 n introduces a need for exploratory analysis to reveal and understand hidd
 en structure in the collected (high-dimensional) Big Data. One crucial asp
 ect in enabling such analysis\, especially in fields with few domain exper
 ts\, is to produce reliable\, robust\, and human-interpretable visualizati
 ons that emphasize desired trends in the data. In this talk\, I will appro
 ach this goal by combining together kernel methods and deep learning to ca
 pture clusters and dynamics in data. In particular\, I will focus on laten
 t progression patterns that often exist in modern data (e.g.\, due to natu
 ral development or guided by external stimuli)\, and interpretable charact
 erization of transition pathways within them\, which is crucial in explora
 tory settings. For example\, in genomic and proteomic data analysis\, cell
 s are actively differentiating or progressing in response to signals\, and
  characterizing these progressions can unlock deep understanding of normal
  development\, as well as enable detection of abnormal transitions (e.g.\,
  cancerous metastasis). To provide such analysis\, I will present PHATE (P
 otential of Heat-diffusion for Affinity-based Transition Embedding) - a no
 vel unsupervised low-dimensional embedding for visualization of data\, whi
 ch reveals and emphasizes transitions and emergent progression patterns. T
 his method uses heat diffusion processes to construct an intrinsic data ge
 ometry and compute distances using their free energy potential. The constr
 ucted diffusion-potential geometry captures high-dimensional transition st
 ructures (when they exist) while enabling their visualization via a low-di
 mensional embedding that approximates local and global nonlinear relations
  in the data. The effectiveness of the produced visualization for explorat
 ory data analysis will be demonstrated on both synthetic and real data\, i
 ncluding facial expressions and new scRNA-seq data of embryoid body develo
 pment that was collected specifically to support development and validatio
 n of this method. Finally\, I will discuss future directions for advancing
  deep learning tools in exploratory settings based on the principles enabl
 ed by these developments.\n	\n	Monsieur Wolf est candidat pour un poste en a
 pprentissage automatique au Département de mathématiques et de statistique
 .\n
DTSTART:20180424T143000Z
DTEND:20180424T153000Z
LOCATION:Room 5340\, CA\, Pav. André-Aisenstadt\, 2920\, ch. de la Tour
SUMMARY:Guy Wolf\, Yale University
URL:https://www.mcgill.ca/mathstat/channels/event/guy-wolf-yale-university-
 286738
END:VEVENT
END:VCALENDAR