Machine Learning Methods for Small Data Challenges in Molecular Science
Методы машинного обучения для решения задач, связанных с малыми объемами данных в молекулярной науке
2023-06-29
SCID: 54.1/udfr7dms
Discuss with AI
data augmentationdeep learningmachine learningmolecular sciencesmall data challenges
Figures from the paper
Abstract (AI)
Small data are often used in scientific and engineering research due to the presence of various constraints, such as time, cost, ethics, privacy, security, and technical limitations in data acquisition. However, big data have been the focus for the past decade, small data and their challenges have received little attention, even though they are technically more severe in machine learning (ML) and deep learning (DL) studies. Overall, the small data challenge is often compounded by issues, such as data diversity, imputation, noise, imbalance, and high-dimensionality. Fortunately, the current big data era is characterized by technological breakthroughs in ML, DL, and artificial intelligence (AI), which enable data-driven scientific discovery, and many advanced ML and DL technologies developed for big data have inadvertently provided solutions for small data problems. As a result, significant progress has been made in ML and DL for small data challenges in the past decade. In this review, we summarize and analyze several emerging potential solutions to small data challenges in molecular science, including chemical and biological sciences. We review both basic machine learning algorithms, such as linear regression, logistic regression (LR), k -nearest neighbor (KNN), support vector machine (SVM), kernel learning (KL), random forest (RF), and gradient boosting trees (GBT), and more advanced techniques, including artificial neural network (ANN), convolutional neural network (CNN), U-Net, graph neural network (GNN), Generative Adversarial Network (GAN), long short-term memory (LSTM), autoencoder, transformer, transfer learning, active learning, graph-based semi-supervised learning, combining deep learning with traditional machine learning, and physical model-based data augmentation. We also briefly discuss the latest advances in these methods. Finally, we conclude the survey with a discussion of promising trends in small data challenges in molecular science.
Key Findings
1
Combining machine learning with domain knowledge and physical models is identified as a promising direction for future molecular-science applications.
2
Methods originally developed for big-data machine learning and deep learning have enabled substantial progress in addressing small-data problems.
3
Small datasets create severe machine-learning challenges involving limited diversity, missing values, noise, class imbalance, and high dimensionality.
4
Small-data constraints in molecular science arise from time, cost, ethical, privacy, security, and technical limitations in data acquisition.
5
The review analyzes classical algorithms and advanced approaches, including transfer learning, active learning, semi-supervised learning, generative models, and physics-based data augmentation.
Research Object
small-data problems in molecular science, including chemical and biological sciences
Research Subject
machine-learning and deep-learning solutions for handling data diversity, imputation, noise, imbalance, and high dimensionality in molecular-science small-data applications
Publication Details
Publication Date
2023-06-29
Journal
Publisher
ISSN
Cited by
503
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai20
ImageNet classification with deep convolutional neural networks2017
Very Deep Convolutional Networks for Large-Scale Image Recognition2014
Gapped BLAST and PSI-BLAST: a new generation of protein database search programs1997
Gradient-based learning applied to document recognition1998
Highly accurate protein structure prediction with AlphaFold2021
Semi-Supervised Classification with Graph Convolutional Networks2016
Greedy function approximation: A gradient boosting machine.2001
A Comprehensive Survey on Graph Neural Networks2020
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks2015
Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups2012
Machine learning: Trends, perspectives, and prospects2015
The Graph Neural Network Model2008
Transformers: State-of-the-Art Natural Language Processing2020
Dynamic Graph CNN for Learning on Point Clouds2019
A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures2019
Image Style Transfer Using Convolutional Neural Networks2016
Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing2022
Generative Adversarial Networks: An Overview2018
Convolutional neural networks: an overview and application in radiology2018
Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules2018