Random Structure of Error Surfaces: Toward New Stochastic Learning Methods Invited Presentation
نویسنده
چکیده
Learning in neural networks can be formulated as global optimization of a multimodal error function that is defined over the high-dimensional space of connection weights. This global optimization is both theoretically intractable [26] [35] and difficult in practice. Traditional learning heuristics, e.g., back-propagation [31] or Boltzmann learning [11], are largely based on gradient methods or stochastic hill-climbing, reflecting traditional global optimization approaches (see, e.g., [10] for a survey). While these methods have shown promise on smaller problem instances, optimizing larger connectionist architectures raises several difficulties which motivate the present work:
منابع مشابه
Stochastic DEA with Using of Skew-Normal Distribution in Error Structure
The stochastic data envelopment analysis (SDEA) was developed considering the value ofinputs and outputs as random variables. Therefore, statistical distributions play an importantrole in this regard. The skew-normal (SN) distribution is a family of probability densityfunctions that is frequently used in practical situations. In this paper, we assume that the inputand output variables are skew-...
متن کاملStochastic First-Order Methods with Random Constraint Projection
We consider convex optimization problems with structures that are suitable for sequential treatment or online sampling. In particular, we focus on problems where the objective function is an expected value, and the constraint set is the intersection of a large number of simpler sets. We propose an algorithmic framework for stochastic first-order methods using random projection/proximal updates ...
متن کاملPractical Anatomy as an Advance Organizer for Anatomy Lectures: Effectiveness in Learning Facilitation for Dental Students
Introduction: Several approaches are introduced to facilitate anatomy learning. One of them is to present practical course before the theoretical one. This study compared the effect of presentation of practical class before lecture on enhancement of learning among dental students of Babol University of Medical Sciences. Methods: Thirty eight students of first semester in school of dentistry of...
متن کاملCluster-Based Image Segmentation Using Fuzzy Markov Random Field
Image segmentation is an important task in image processing and computer vision which attract many researchers attention. There are a couple of information sets pixels in an image: statistical and structural information which refer to the feature value of pixel data and local correlation of pixel data, respectively. Markov random field (MRF) is a tool for modeling statistical and structural inf...
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 2004