Lecturer : Rob Schapire Lecture # 9
نویسنده
چکیده
However, suppose a learning algorithm cannot attain error rates less than some fixed amount, say 40%. Is it possible, despite this limitation, to still drive the error rate arbitrarily close to zero? This question motivates the weak learning model, which will be described below. This model will consider circumstances in which the learning algorithms can only do slightly better than random guessing, that is, error probability less than 50%. A concept class C is weakly learnable if ∃ γ > 0 ∃ algorithm A ∀ c ∈ C ∀ D ∀ δ > 0 A takes m = poly ( 1 δ , · · · ) examples to compute a hypothesis h such that
منابع مشابه
An Automatic Lecture Recording System Using Pan-Tilt-Zoom Camera to Track Lecturer and Handwritten Data
For automatic recording of the traditional lecture with a wide blackboard by using a Pan-Tilt-Zoom (PTZ) camera, a system, called ACRS (After Class Reviewing System), was proposed. From the point of view of students reviewing the lecture, the handwritten data is more important than the video of the lecturer. So, the handwritten data and the lecturer were considered together to record a lecture ...
متن کاملOptimization II Winter 2009 / 10 Lecture 5 — November 9
In this lecture we present a procedure which can be thought of as a derandomization of the randomized fictitious play described in the previous lecture. Freund and Schapire [FS99] used this method, which was originally developed by Littlestone and Warmuth [LW94] from the machine learning community, to give a procedure for computing εsaddle points for matrix games. A number of similar algorithms...
متن کاملLecture Capture Using Large Interactive Display Systems
There are various software technologies that allow capture and redelivery of lectures. Most of these technologies however rely on the use of proprietary software, often requiring extra efforts from the lecturer in terms of the initial preparation of the lecture material, or in editing and annotating after the lecture to make the material suitable for the students. To review the material student...
متن کاملAnthropocentric Video Segmentation for Lecture Webcasts
Many lecture recording and presentation systems transmit slides or chalkboard content along with a small video of the instructor. As a result, two areas of the screen are competing for the viewer’s attention, causing the widely known split-attention effect. Face and body gestures, such as pointing, do not appear in the context of the slides or the board. To eliminate this problem, this article ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2008