Demystifying Deep Learning: A Geometric Approach to Iterative Projections
نویسندگان
چکیده
Parametric approaches to Learning, such as deep learning (DL), are highly popular in nonlinear regression, in spite of their extremely difficult training with their increasing complexity (e.g. number of layers in DL). In this paper, we present an alternative semi-parametric framework which foregoes the ordinarily required feedback, by introducing the novel idea of geometric regularization. We show that certain deep learning techniques such as residual network (ResNet) architecture are closely related to our approach. Hence, our technique can be used to analyze these types of deep learning. Moreover, we present preliminary results which confirm that our approach can be easily trained to obtain complex structures.
منابع مشابه
Curvature-based Comparison of Two Neural Networks
In this paper we show the similarities and differences of two deep neural networks by comparing the manifolds composed of activation vectors in each fully connected layer of them. The main contribution of this paper includes 1) a new data generating algorithm which is crucial for determining the dimension of manifolds; 2) a systematic strategy to compare manifolds. Especially, we take Riemann c...
متن کاملThe Relationship of Study and Learning approaches with Students’ Academic Achievement in Rafsanjan University of Medical Sciences
Introduction: Most experts consider learning approach as the fundamental basis of learning dividing it into two parts of deep learning approach and surface learning approach. This is an endeavor to investigate the relationship between learning and study approaches with academic achievement among students in Rafsanjan University of Medical Sciences. Methods: This descriptive cross-sectional stu...
متن کاملتأثیر آموزش راهبردهای خود تنظیمی بر رویکردهای یادگیری دانش آموزان اول دبیرستان
Abstract The present study was conducted to determine the effect of learning self-regulation strategies on surface, deep and strategic learning approaches of high school first grade female students in Yazd. The study method was pre-test and post-test design. For this purpose, a sample size of 57 subjects was selected by multistage cluster sampling method among high school first grade female ...
متن کاملDetection of children's activities in smart home based on deep learning approach
Monitoring behavior of children in the home is the extremely important to avoid the possible injuries. Therefore, an automated monitoring system for monitoring behavior of children by researchers has been considered. The first step for designing and executing an automated monitoring system on children's behavior in closed spaces is possible with recognize their activity by the sensors in the e...
متن کاملA model based, anatomy dependent method for ultra-fast creation of primary SPECT projections
Introduction: Monte Carlo (MC) is the most common method for simulating virtual SPECT projections. It is useful for optimizing procedures, evaluating correction algorithms and more recently image reconstruction as a forward projector in iterative algorithms; however, the main drawback of MC is its long run time. We introduced a model based method considering the eff...
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 2018