Transfer-RLS method and transfer-FORCE learning for simple and fast training of reservoir computing models
Transfer-RLS method and transfer-FORCE learning for simple and fast training of reservoir computing models
复制标题
Transfer-RLS方法和transfer-FORCE学习用于简单快速地训练油藏计算模型
DOI：
10.1016/j.neunet.2021.06.031
复制
发表时间：
2021
期刊：
影响因子：
7.8
通讯作者：
Hiroto Tamura and Gouhei Tanaka
中科院分区：
文献类型：
--
作者：
Maniruzzaman Md.;Hasan Md. Al Mehedi;Asai Nobuyoshi;Shin Jungpil;Hiroto Tamura and Gouhei Tanaka
油藏计算是一种由一种特殊类型的递归神经网络衍生而来的机器学习框架。随着物理油藏计算的最新进展，一些油藏计算设备被认为是用于实时信息处理的节能机器学习硬件。这有利于开发快速收敛、操作简单的学习方法，利用低功耗的油藏计算设备实现高效的在线学习。本文提出了一种介于递归最小二乘(RLS)方法和最小均方(LMS)方法之间的训练方法，这两种方法是水库计算模型的标准在线学习方法。RLS方法收敛速度快，但需要更新称为增益矩阵的巨大矩阵，而LMS方法不使用增益矩阵，但收敛非常慢。另一方面，所提出的方法称为Transfer-RLS方法，不需要在主训练阶段(即，在预训练阶段)通过提前更新增益矩阵来更新增益矩阵。结果表明，与原RLS方法相比，Transfer-RLS方法具有更简单的运算，且不牺牲太多的收敛速度。此外，我们证明了一种改进的传递-RLS方法(称为传递-力学习)可以应用于具有闭环系统的油藏计算模型的一阶降阶和受控(力)学习，这是一种具有挑战性的训练。数值计算和分析表明，Transfer-RLS方法的收敛速度比LMS方法快得多。
Reservoir computing is a machine learning framework derived from a special type of recurrent neural network. Following recent advances in physical reservoir computing, some reservoir computing devices are thought to be promising as energy-efficient machine learning hardware for real-time information processing. It is beneficial to develop fast convergence learning methods with simpler operations, to realize efficient online learning with low-power reservoir computing devices. This study proposes a training method located in the middle between the recursive least squares (RLS) method and the least mean squares (LMS) method, which are standard online learning methods for reservoir computing models. The RLS method converges fast but requires updates of a huge matrix called a gain matrix, whereas the LMS method does not use a gain matrix but converges very slow. On the other hand, the proposed method called a transfer-RLS method does not require updates of the gain matrix in the main-training phase by updating that in advance (i.e., in a pre-training phase). As a result, the transfer-RLS method can work with simpler operations than the original RLS method without sacrificing much convergence speed. Besides, we show that a modified version of the transfer-RLS method (called transfer-FORCE learning) can be applied to the first-order reduced and controlled (FORCE) learning for a reservoir computing model with a closed-loop, which is challenging to train. Furthermore, we numerically and analytically show that the transfer-RLS method converges much faster than the LMS method.