Unsupervised and semi‐supervised learning: the next frontier in machine learning for plant systems biology

Abstract

Advances in high-throughput omics technologies are leading plant biology research into the era of big data. Machine learning (ML) performs an important role in plant systems biology because of its excellent performance and wide application in the analysis of big data. However, to achieve ideal performance, supervised ML algorithms require large numbers of labeled samples as training data. In some cases, it is impossible or prohibitively expensive to obtain enough labeled training data; here, the paradigms of unsupervised learning (UL) and semi-supervised learning (SSL) play an indispensable role. In this review, we first introduce the basic concepts of ML techniques, as well as some representative UL and SSL algorithms, including clustering, dimensionality reduction, self-supervised learning (self-SL), positive-unlabeled (PU) learning and transfer learning. We then review recent advances and applications of UL and SSL paradigms in both plant systems biology and plant phenotyping research. Finally, we discuss the limitations and highlight the significance and challenges of UL and SSL strategies in plant systems biology.

A review of artificial intelligence-assisted omics techniques in plant defense: current trends and future directions

“…Ensemble classifiers enhance decision-making by combining outputs from different models, introduced by Dietterich (2000) . Clustering-based methods, like the k-means algorithm, provide an unsupervised approach for predicting protein functions by exploiting direct and indirect interactions ( Hou, 2017 ; Yan and Wang, 2022 ).…”

Meta‐learning shows great potential in plant disease recognition under few available samples

“…Finally, there is the need to consider how meta‐learning techniques can be used to assist in micro‐level research (Mason et al., 2021). Genomic and proteomic data at multiple levels within plant systems are characterized by high latitude, redundancy, noise, uncertainty, and multiple sources (Yan & Wang, 2022), making them quite difficult to study. Some studies have attempted to use unsupervised and semi‐supervised learning to assist in the exploration and integration of knowledge with decent results.…”

An Outlook for Deep Learning in Ecosystem Science

“…Various solutions have been proposed to address the issue of limited training data based on concerns that models trained on small datasets are vulnerable to overfitting. First, although the large majority of ecological applications use supervised learning, the development of unsupervised and self-supervised algorithms that circumvent the need for extensive labelled training data is an active area of research (for example, Yan and Wang 2022).…”