Sharing Residual Units Through Collective Tensor Factorization in Deep Neural Networks

نویسندگان

  • Yunpeng Chen
  • Xiaojie Jin
  • Bingyi Kang
  • Jiashi Feng
  • Shuicheng Yan
چکیده

Residual units are wildly used for alleviating optimization difficulties when building deep neural networks. However, the performance gain does not well compensate the model size increase, indicating low parameter efficiency in these residual units. In this work, we first revisit the residual function in several variations of residual units and demonstrate that these residual functions can actually be explained with a unified framework based on generalized block term decomposition. Then, based on the new explanation, we propose a new architecture, Collective Residual Unit (CRU), which enhances the parameter efficiency of deep neural networks through collective tensor factorization. CRU enables knowledge sharing across different residual units using shared factors. Experimental results show that our proposed CRU Network demonstrates outstanding parameter efficiency, achieving comparable classification performance to ResNet-200 with the model size of ResNet-50. By building a deeper network using CRU, we can achieve stateof-the-art single model classification accuracy on ImageNet-1k and Places365-Standard benchmark datasets. (Code and trained models are available on GitHub3)

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Global Optimality in Tensor Factorization, Deep Learning, and Beyond

Techniques involving factorization are found in a wide range of applications and have enjoyed significant empirical success in many fields. However, common to a vast majority of these problems is the significant disadvantage that the associated optimization problems are typically non-convex due to a multilinear form or other convexity destroying transformation. Here we build on ideas from conve...

متن کامل

Wide Compression: Tensor Ring Nets

Deep neural networks have demonstrated state-of-theart performance in a variety of real-world applications. In order to obtain performance gains, these networks have grown larger and deeper, containing millions or even billions of parameters and over a thousand layers. The tradeoff is that these large architectures require an enormous amount of memory, storage, and computation, thus limiting th...

متن کامل

Neuron Mathematical Model Representation of Neural Tensor Network for RDF Knowledge Base Completion

In this paper, a state-of-the-art neuron mathematical model of neural tensor network (NTN) is proposed to RDF knowledge base completion problem. One of the difficulties with the parameter of the network is that representation of its neuron mathematical model is not possible. For this reason, a new representation of this network is suggested that solves this difficulty. In the representation, th...

متن کامل

From Deep to Shallow: Transformations of Deep Rectifier Networks

In this paper, we introduce transformations of deep rectifier networks, enabling the conversion of deep rectifier networks into shallow rectifier networks. We subsequently prove that any rectifier net of any depth can be represented by a maximum of a number of functions that can be realized by a shallow network with a single hidden layer. The transformations of both deep rectifier nets and deep...

متن کامل

Cystoscopy Image Classication Using Deep Convolutional Neural Networks

In the past three decades, the use of smart methods in medical diagnostic systems has attractedthe attention of many researchers. However, no smart activity has been provided in the eld ofmedical image processing for diagnosis of bladder cancer through cystoscopy images despite the highprevalence in the world. In this paper, two well-known convolutional neural networks (CNNs) ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • CoRR

دوره abs/1703.02180  شماره 

صفحات  -

تاریخ انتشار 2017