Advanced methods for optimizing work with neural networks on modern architectures
Journal Title: Modern Innovations, Systems and Technologies - Year 2024, Vol 4, Issue 4
Abstract
The paper is devoted to the study and comparison of different optimization methods for training neural networks. Key algorithms such as stochastic gradient descent, Momentum method, AdaGrad and Adam are considered. For each method, theoretical justifications, mathematical formulas and examples of practical implementation are provided. An experimental comparison of the performance of these methods on the task of handwritten digit classification using the MNIST dataset is carried out. The advantages and disadvantages of each method are discussed, as well as their impact on the learning rate and accuracy of the model. Based on the results obtained, the most efficient algorithm is summarized, and the importance of selecting an appropriate optimization method to improve the efficiency of neural networks in various applications is emphasized.
Authors and Affiliations
P. M. Urvachev, V. A. Kovtun
Keywords
Related Articles
- EP ID EP755369
- DOI 10.47813/2782-2818-2024-4-4-0199-0212
- Views 25
- Downloads 0
Using the least squares method to find predicted values
The least squares method is used to analyze experimentally obtained data. The purpose of this method is to find a function that is closest to the results obtained in practice. This method makes sense to use when construc...
Contact problems of the theory of roller squeezing of leather
The article deals with the main contact problems of the theory of roller squeezing of leather. Mathematical models of the shape of the roll contact curves, of friction stresses and the regularities of the distribution o...
Security of human private life when interacting with generative artificial intelligence
This scholarly article critically examines the intricate dynamics of personal privacy in the age of pervasive internet use, emphasizing the escalating challenges posed by Generative Artificial Intelligence. Every year th...
Influence of raw properties on product quality
The purpose of this article is to study methods for determining and predicting the drapability of a fabric. Drapability is one of the main aesthetic properties of fabric that must be taken into account when designing fab...
Modeling the water cycle in a bio-regenerative life support system
The work is devoted to modeling the water cycle in the bio-regenerative life support system. The scheme of closed hydrological cycle, including human and plant links, modules of food preparation, waste treatment, water a...