NoAS-DS: Neural optimal architecture search for detection of diverse DNA signals

Long short-term memory with activation on gradient

As the number of long short-term memory (LSTM) layers increases, vanishing/exploding gradient problems exacerbate and have a negative impact on the performance of the LSTM. In addition, the ill-conditioned problem occurs in the training process of LSTM and adversely affects its convergence. In this work, a simple and effective method of the gradient activation is applied to the LSTM, while empirical criteria for choosing gradient activation hyperparameters are found. Activating the gradient refers to modifying the gradient with a specific function named the gradient activation function. Moreover, different activation functions and different gradient operations are compared to prove that the gradient activation is effective on LSTM. Furthermore, comparative experiments are conducted, and their results show that the gradient activation alleviates the above problems and accelerates the convergence of the LSTM. The source code is publicly available at https://github.com/LongJin-lab/ACT-In-NLP.

Ensemble recurrent neural network with whale optimization algorithm-based DNA sequence classification for medical applications

The modern data-driven era has facilitated the gathering of large quantities of biomedical and clinical data. The deoxyribonucleic acid gene expression datasets have become a vital focus for the research community because of their capability to detect pathogens via 'biomarkers' or particular modifications in the gene sequence which portray a specific pathogen. Metaheuristic-related feature selection (FS) efficiently filters out only the pertinent genes out of large feature sets to lessen the data storage and computation requirements. This paper embraces the whale optimization algorithm for the FS issue in HD microarray data for the effectual propagation of candidate solutions to reach global optima over sufficient iterations. The chosen data are classified by employing an ensemble recurrent network (ERNN) that retains the amalgamation of long short-term memory, bidirectional long short-term memory, and gated recurrent units. Analysis of this proposed ERNN methodology would be performed by correlating with diverse advanced methodologies, and thus, the ERNN attains 99.59% precision and 99.59% accuracy.