A review on soft computing approaches for predicting maintainability of software: State‐of‐the‐art, technical challenges, and future directions

Abstract

The software is changing rapidly with the invention of advanced technologies and methodologies. The ability to rapidly and successfully upgrade software in response to changing business requirements is more vital than ever. For the long‐term management of software products, measuring software maintainability is crucial. The use of soft computing techniques for software maintainability prediction has shown immense promise in software maintenance process by providing accurate prediction of software maintainability. To better understand the role of soft computing techniques for software maintainability prediction, we aim to provide a systematic literature review of soft computing techniques for predicting software maintainability. Firstly, we provide a detailed overview of software maintainability. Following this, we explore the fundamentals of software maintainability and the reasons for adopting soft computing methodologies for predicting software maintainability. Later, we examine the soft computing approaches employed in the process of software maintainability prediction. Furthermore, we discuss the difficulties and potential solutions associated with the use of soft computing techniques in predicting maintainability of software. Finally, we conclude the review with some promising future directions to drive further research innovations and developments in this promising area. This systematic literature review provides a comprehensive overview of the soft computing strategies utilized for software maintainability for future researchers.

OmniNA: A foundation model for nucleotide sequences

“…We replace the ReLU non-linearity of the traditional transformer architecture by SwiGLU activation function 16 . The output of the last transformer layer is fed into a linear layer for auto-regressive learning 5 . The sequence and annotations were trained in an auto-regression.…”

“…The input underwent chunking into subword tokens, followed by embedding through the token embedding layer. Token embeddings were then concatenated with position embeddings and fed into the transformer layers and followed by a regression layer for auto-regressive training (Figure 1A and Supplementary Figure 2) 5 . The pre-trained OmniNA can handle versatile downstream applications in a multi-task manner (Figure 1B).…”

“…These models learn the underlying data patterns and structures from massive unlabeled data, which facilitate innovative applications via transfer learning [2][3][4] . It has emerged as a transformative technology with many successful examples in various fields, such as the Generative Pre-trained Transformer (GPT) 5 in natural language processing and Stable Diffusion 6 in vision language. In the ever-evolving landscape of genomics, the advent of the foundation model has emerged as a transformative force, propelling our capacity to decode the intricate language embedded within nucleic acid sequences.…”

A Review of Load Frequency Control Schemes Deployed for Wind-Integrated Power Systems

“…Thus, the growing recognition of the importance of frequency regulation in renewable and conventional power networks has resulted in a particular focus on soft computing techniques. The concept of soft computing, in contrast to hard computing, allows for implicit assumptions, imprecisions, ambiguities, and partial truths [94]. In recent years, soft computing, which is inspired by human brains, has become a popular research and study topic within the LFC scheme of power systems [95,96].…”