Meta Networks

نویسندگان

  • Tsendsuren Munkhdalai
  • Hong Yu
چکیده

Deep neural networks have been successfully applied in applications with a large amount of labeled data. However, there are major drawbacks of the neural networks that are related to rapid generalization with small data and continual learning of new concepts without forgetting. We present a novel meta learning method, Meta Networks (MetaNet), that acquires a meta-level knowledge across tasks and shifts its inductive bias via fast parameterization for the rapid generalization. When tested on the standard oneshot learning benchmarks, our MetaNet models achieved near human-level accuracy. We demonstrated several appealing properties of MetaNet relating to generalization and continual learning.

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

Modeling Structural Relationships of Meta-Cognitive Situations with Tendency to Virtual Networks through Mediating of Emotional Processing in Gifted Students

Background: Cyberspace covers many aspects of human life and tendency to cyberspace can be influenced by cognitive and emotional dimensions. Thus, the aim of this study was modeling structural relationships of meta-cognitive situations with tendency for virtual networks through mediating emotional processing in gifted students. Methods: The research method was descriptive-correlation and in pa...

متن کامل

A meta-heuristic clustering method to reduce energy consumption in Internet of Things

The Internet of Things (IoT) is an emerging phenomenon in the field of communication, in which smart objects communicate with each other and respond to user requests. The IoT provides an integrated framework providing interoperability across various platforms. One of the most essential and necessary components of IoT is wireless sensor networks. Sensor networks play a vital role in the lowest l...

متن کامل

Combining Harmony search algorithm and Ant Colony Optimization algorithm to increase the lifetime of Wireless Sensor Networks

Wireless Sensor Networks are the new generation of networks that typically are formed great numbers of nodes and the communications of these nodes are done as Wireless. The main goal of these networks is collecting data from neighboring environment of network sensors. Since the sensor nodes are battery operated and there is no possibility of charging or replacing the batteries, the lifetime of ...

متن کامل

Presenting a comprehensive framework of effective factors, behavior and consequences of using enterprise social networks

Aim: The purpose of this paper is to determine the factors affecting the use of enterprise social networks, the types of usage behavior and the consequences of using these networks. Methodology: This is a descriptive study. The qualitative inquiries of previous researches in the social networking enterprise had been investigated by means of  meta-synthesis. A total of 470 found source of meta-s...

متن کامل

Iran's Electrical Energy Demand Forecasting Using Meta-Heuristic Algorithms

This study aims to forecast Iran's electricity demand by using meta-heuristic algorithms, and based on economic and social indexes. To approach the goal, two strategies are considered. In the first strategy, genetic algorithm (GA), particle swarm optimization (PSO), and imperialist competitive algorithm (ICA) are used to determine equations of electricity demand based on economic and social ind...

متن کامل

A HYBRID CHARGED SYSTEM SEARCH - FIREFLY ALGORITHM FOR OPTIMIZATION OF WATER DISTRIBUTION NETWORKS

Water distribution networks are one of the important and costly infrastructures of cities and many meta-heuristic algorithms in standard or hybrid forms were used for optimizing water distribution networks. These algorithms require a large amount of computational cost. Therefore, the converging speed of algorithms toward the optimization goal is as important as the goal itself. In this paper, a...

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:

دوره   شماره 

صفحات  -

تاریخ انتشار 2017