Heavy ball momentum is crucial in accelerating (stochastic) gradient-based optimization algorithms for machine learning. Existing heavy ball momentum is usually weighted by a uniform hyperparameter, which relies on excessive tuning. Moreover, the calibrated fixed hyperparameter may not lead to optimal performance. In this paper, to eliminate the effort for tuning the momentum-related hyperparameter, we propose a new adaptive momentum inspired by the optimal choice of the heavy ball momentum for quadratic optimization. Our proposed adaptive heavy ball momentum can improve stochastic gradient descent (SGD) and Adam. SGD and Adam with the newly designed adaptive momentum are more robust to large learning rates, converge faster, and generalize better than the baselines. We verify the efficiency of SGD and Adam with the new adaptive momentum on extensive machine learning benchmarks, including image classification, language modeling, and machine translation. Finally, we provide convergence guarantees for SGD and Adam with the proposed adaptive momentum.
The voltage is unstable due to the influence of load demand and other factors. In order to ensure the stable operation of the power system, this paper proposes a method of power system stability margin prediction based on a gradient lifting decision tree. On this basis, the stable sampling of power grid is established by using data mining method and solved. On this basis, combined with the stability simulation curve of the power grid, the operation characteristics of the power grid are analyzed, and the stability of the power grid is analyzed using the CNN model and the gradient lifting decision tree. The proposed method has better prediction accuracy and efficiency of power system stability margin, and can improve reference for ensuring the stable operation.
Power information communication contains sensitive and private data, which may be illegally stolen and tampered after being leaked, posing a threat to the data security of power enterprises and users. Aiming at the problem of limited key space, a data leakage prevention method for power information communication based on chaotic mapping is proposed. The physical layer of power communication includes power data produced and distributed by power equipment, and the information layer includes operation, monitoring and dispatching data. There is an association between the two data. The dispatch and control center uses the association relationship to establish access rights for users and generate access private keys for trusted users. According to the user authority, the communication data is encrypted and decrypted by chaotic mapping, so that the two adjacent components of the reconstructed key space are relatively independent, and the properties of the chaotic sequence are maintained. The test results show that the proposed method can effectively reduce the computing and communication overhead of the power data encryption process and improve the encryption efficiency.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.