Fast training of recurrent networks based on the EM algorithm
نویسندگان
چکیده
In this work, a probabilistic model is established for recurrent networks. The expectation-maximization (EM) algorithm is then applied to derive a new fast training algorithm for recurrent networks through mean-field approximation. This new algorithm converts training a complicated recurrent network into training an array of individual feedforward neurons. These neurons are then trained via a linear weighted regression algorithm. The training time has been improved by five to 15 times on benchmark problems.
منابع مشابه
A unified approach on fast training of feedforward and recurrent networks using EM algorithm
In this work, we provide a theoretical framework that unifies the notions of hidden representations and moving targets through the expectation and maximization (EM) algorithm. Based on such a framework, two fast training algorithms can be derived consistently for both feedforward networks and recurrent networks.
متن کاملVery Fast Load Flow Calculation Using Fast-Decoupled Reactive Power Compensation Method for Radial Active Distribution Networks in Smart Grid Environment Based on Zooming Algorithm
Distribution load flow (DLF) calculation is one of the most important tools in distribution networks. DLF tools must be able to perform fast calculations in real-time studies at the presence of distributed generators (DGs) in a smart grid environment even in conditions of change in the network topology. In this paper, a new method for DLF in radial active distribution networks is proposed. The ...
متن کاملGGRA: a grouped gossip-based reputation aggregation algorithm
An important issue in P2P networks is the existence of malicious nodes that decreases the performance of such networks. Reputation system in which nodes are ranked based on their behavior, is one of the proposed solutions to detect and isolate malicious (low ranked) nodes. Gossip Trust is an interesting previously proposed algorithm for reputation aggregation in P2P networks based on t...
متن کاملAn Irregular Lattice Pore Network Model Construction Algorithm
Pore network modeling uses a network of pores connected by throats to model the void space of a porous medium and tries to predict its various characteristics during multiphase flow of various fluids. In most cases, a non-realistic regular lattice of pores is used to model the characteristics of a porous medium. Although some methodologies for extracting geologically realistic irregular net...
متن کاملDynamic Obstacle Avoidance by Distributed Algorithm based on Reinforcement Learning (RESEARCH NOTE)
In this paper we focus on the application of reinforcement learning to obstacle avoidance in dynamic Environments in wireless sensor networks. A distributed algorithm based on reinforcement learning is developed for sensor networks to guide mobile robot through the dynamic obstacles. The sensor network models the danger of the area under coverage as obstacles, and has the property of adoption o...
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:
- IEEE transactions on neural networks
دوره 9 1 شماره
صفحات -
تاریخ انتشار 1998