Calculation of soil type is an important part of the earthwork project, and its result determines the subsequent earth pit operation mode, which in turn determines whether the subsequent civil construction work can be...
详细信息
Neural network algorithm (NNA) is a new type of population-based meta-heuristic algorithm, which is inspired by artificial neural networks. NNA is a result of the deep integration of artificial neural networks and met...
详细信息
In machine learning and neural network optimization, algorithms like incremental gradient, and shuffle SGD are popular due to minimizing the number of cache misses and good practical convergence behavior. However, the...
详细信息
In machine learning and neural network optimization, algorithms like incremental gradient, and shuffle SGD are popular due to minimizing the number of cache misses and good practical convergence behavior. However, their optimization properties in theory, especially for non-convex smooth functions, remain incompletely explored. This paper delves into the convergence properties of SGD algorithms with arbitrary data ordering, within a broad framework for non-convex smooth functions. Our findings show enhanced convergence guarantees for incremental gradient and single shuffle SGD. Particularly if n is the training set size, we improve n times the optimization term of convergence guarantee to reach accuracy Ε from O (n/Ε) to O (1/Ε). Copyright 2024 by the author(s)
作者：
Shao, WeijiaUnit 2.6 Workplaces
Safety of Machinery Operational Safety Federal Institute for Occupational Safety and Health Dresden Germany
This work proposes an algorithm improving the dimensionality dependence for gradient-free optimisation over cross-polytopes, which has many applications such as adversarial attacks, explainable AI and sparse regressio...
详细信息
This work proposes an algorithm improving the dimensionality dependence for gradient-free optimisation over cross-polytopes, which has many applications such as adversarial attacks, explainable AI and sparse regression. For bandit convex optimisation with two-point feedback over cross-polytopes, the state-of-the-art algorithms have a dimensionality dependence of O(√dlog d), while the known lower bound is of the form Ω(pd(log d)−1). We propose a mirror descent algorithm equipped with a symmetric version of the negative 12-Tsallis entropy. Combined with an 1-ellipsoidal smoothing-based gradient estimator, the proposed algorithm guarantees a dimensionality dependence on O(√d), which improves the state-of-the-art algorithms by a factor of √log d. The idea can be further applied to optimising nonsmooth and non-convex functions. We propose an algorithm with a convergence depending on O(d), which is the best-known dimensionality dependence. Copyright 2024 by the author(s)
Extensive research on edge inference has devoted in optimizing service performance for users. However, recent studies have overlooked the desired utility of application service provider (ASP), which is crucial for ach...
详细信息
Hyperparameter optimization on Machine Learning models is crucial for their correct refinement. For complex big models (such as Deep Learning models), in which a single training model is supposed to have a very high c...
详细信息
Objective: This study aims to explore the optimization of XGBoost algorithm parameters based on heuristic algorithms, with the goal of improving the classification accuracy of the ***: Leveraging the principles of bio...
详细信息
In this paper, we re-examine the role of convexity and smoothness on gradient-based unconstrained optimization. While existing literature establishes the fundamental limits for gradient-based optimization algorithms f...
详细信息
Aiming at the current problem of insufficient autonomous flight and obstacle avoidance ability of UAVs in complex environments and poor algorithm optimization effect. In this paper, a multi-angle improved Gray Wolf al...
详细信息
Air overpressure (AOp) induced by rock blasting is an undesirable phenomenon in open-pit mines and civil construction works. The prediction of AOp has been always a complicated task since many parameters have potentia...
详细信息
Air overpressure (AOp) induced by rock blasting is an undesirable phenomenon in open-pit mines and civil construction works. The prediction of AOp has been always a complicated task since many parameters have potential to affect the propagation of air waves. This study aims to assess the capability of a new hybrid evolutionary model based on an integrated adaptive neuro-fuzzy inference system (ANFIS) with a stochastic fractal search (SFS) algorithm. To assess the reliability and acceptability of ANFIS-SFS model, the particle swarm optimization (PSO) and genetic algorithm (GA) were also combined with ANFIS. The proposed models were developed using a comprehensive database including 62 sets of data collected from four granite quarry sites in Malaysia. Performances of the ANFIS-SFS, ANFIS-GA, and ANFIS-PSO models were checked using statistical functions as the performance criteria. The obtained results showed that the proposed ANFIS-SFS model, with root mean square error of 1.223 dB, provided much higher generalization capacity than the ANFIS-PSO (RMSE of 1.939 dB), ANFIS-GA (RMSE of 2.418 dB), and ANFIS (RMSE of 3.403 dB) models in terms of predicting AOp. This clearly demonstrates the effectiveness of SFS to provide a more accurate model in the AOp prediction field.
暂无评论