Optimizing echo state network through a novel fisher maximization based stochastic gradient descent

Öztürk, Muhammed Maruf; Çankaya, İbrahim Arda; İpekçi, Deniz

doi:10.1016/j.neucom.2020.07.034

Cited by 14 publications

(2 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar to the simplest artificial neural network, traditional reservoir computing comprises three basic parts 50 : the input layer, intermediate layer (reservoir), and output layer, as illustrated in Fig. 2 a.…”

Section: Methodsmentioning

confidence: 99%

Short-term memory capacity analysis of Lu3Fe4Co0.5Si0.5O12-based spin cluster glass towards reservoir computing

Liao

Yamahara

Terao

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Reservoir computing is a brain heuristic computing paradigm that can complete training at a high speed. The learning performance of a reservoir computing system relies on its nonlinearity and short-term memory ability. As physical implementation, spintronic reservoir computing has attracted considerable attention because of its low power consumption and small size. However, few studies have focused on developing the short-term memory ability of the material itself in spintronics reservoir computing. Among various magnetic materials, spin glass is known to exhibit slow magnetic relaxation that has the potential to offer the short-term memory capability. In this research, we have quantitatively investigated the short-term memory capability of spin cluster glass based on the prevalent benchmark. The results reveal that the magnetization relaxation of Co, Si-substituted Lu3Fe5O12 with spin glass behavior can provide higher short-term memory capacity than ferrimagnetic material without substitution. Therefore, materials with spin glass behavior can be considered as potential candidates for constructing next-generation spintronic reservoir computing with better performance.

show abstract

Section: Methodsmentioning

confidence: 99%

Short-term memory capacity analysis of Lu3Fe4Co0.5Si0.5O12-based spin cluster glass towards reservoir computing

Liao

Yamahara

Terao

et al. 2023

Sci Rep

View full text Add to dashboard Cite

show abstract

“…However, the second stage is optimization of the nonlinear functional. To find the optimal solution, we use gradient descent [27,28] and penalty functions [29][30][31]. Then problem (10,18,22,23,28,29) takes the form (18,22,23,(30)(31)(32)(33)(34)36) with the iteration rule: The stopping criterion will be considered the achievement of such a value of k so that the inequality Let us consider formulas (30)(31)(32)(33)(34)(35)) in more detail.…”

Section: Stagementioning

confidence: 99%

A mathematical model for the formation of the pricing policy and the plan of the production and transport system in a timber-processing enterprise

Rogulin¹

2021

Bus. Inform.

View full text Add to dashboard Cite

The formation of supply chains for raw materials is closely related to production problems involving the determination of prices for sold goods. The question often arises about the need to study the sources of raw materials and the methodology for pricing the goods produced, taking into account a large number of external aspects of the market. Often, only particular approaches to solving production problems are considered in the literature, and methods for solving the complex problem of forming supply chains for raw materials and pricing are poorly developed. This paper presents a mathematical model that makes it possible to assess the feasibility of interaction between a timber industry enterprise and a commodity exchange, with the daily formation of a price vector over the entire planning horizon. A two-stage algorithm for finding a suboptimal solution is considered, which at the first stage is based on linear optimization, and at the second, on gradient descent with the use of penalty functions. The model was tested on the data of the commodity and raw materials exchange of Russia and one of the enterprises of the Primorsky Territory. The result of testing was the volume of production of each type of product over the entire planning horizon, the volume of delivery of raw materials from regions to enterprises, as well as the methods of delivery of goods to the consumer and the policy of pricing. It is shown that almost all goods should increase in price due to a reduction in the excess volume of applications (demand) over the entire planning horizon, with the exception of two types of products. It is noted that the exchange can provide the necessary volume of raw materials for high-capacity production, which demonstrates the possibility, if necessary, to increase the volume of raw materials purchases. It is shown which goods will be included in the release plan more often than others when optimizing the price vector. The ways of delivery of final types of products are analyzed. The disadvantages and advantages of the mathematical model and algorithm are presented.

show abstract

On the post-hoc explainability of deep echo state networks for time series forecasting, image and video classification

Arrieta

Gil-López

Laña

et al. 2021

Neural Comput & Applic

View full text Add to dashboard Cite

Optimizing echo state network through a novel fisher maximization based stochastic gradient descent

Cited by 14 publications

References 34 publications

Short-term memory capacity analysis of Lu3Fe4Co0.5Si0.5O12-based spin cluster glass towards reservoir computing

Short-term memory capacity analysis of Lu3Fe4Co0.5Si0.5O12-based spin cluster glass towards reservoir computing

A mathematical model for the formation of the pricing policy and the plan of the production and transport system in a timber-processing enterprise

On the post-hoc explainability of deep echo state networks for time series forecasting, image and video classification

Contact Info

Product

Resources

About