BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Karen Seidel (Hasso Plattner Institute\, University of Potsdam) DTSTART:20211130T200000Z DTEND:20211130T210000Z DTSTAMP:20260423T005739Z UID:CTA/71 DESCRIPTION:Title: Mod elling binary classification\nby Karen Seidel (Hasso Plattner Institut e\, University of Potsdam) as part of Computability theory and application s\n\n\nAbstract\nIn machine learning\, algorithms generalise from availabl e training data to unseen situations.\nThe engineering practices used in t he respective technologies are far from understood.\nResearch in \\emph{in ductive inference} analyses concrete mathematical models for this complex subject with tools from computability theory.\n\nWe investigate models for incremental binary classification\, an example for supervised online lear ning.\nOur starting point is a model for human and machine learning sugges ted by E.~M.~Gold.\n\nFor learning algorithms that use all of the availabl e binary labeled training data in order to compute the current hypothesis\ , we observe that the distribution of the training data does not influence learnability.\nWhen approximating the concept to be learned\, we obtain a strict hierarchy\, depending on the error parameter.\nWe consider differe nt hypothesis spaces for spam problem like and symmetric classification ta sks and provide the complete maps.\n LOCATION:https://researchseminars.org/talk/CTA/71/ END:VEVENT END:VCALENDAR