2021 Hot Topics in Machine Learning Research

By: K. Yadav, M. Quamara, B. Gupta

1. Federated Machine Learning
   • What is it about?
     • Federated machine learning is about training a model or an algorithm over dataset across decentralized edge devices in distributed networks via several training rounds until the model or the algorithm converge [1]. Unlike traditional machine learning, federated machine learning introduces a learning paradigm that does not aggregate the dataset from the devices to a single server.
   • Why to use Federated Machine Learning instead of Centralized Machine Learning?
     • Privacy
     • Data breach
     • Real time dataset
     • Peer-to-peer learning
   • Active research areas:
     • Secure gradient sharing
     • Communication rounds
     • Independent and Identically Distributed (IID) and non-IID data
     • Generalization and personalization scheme
     • Adversarial federated learning
     • Resource allocation strategies
     • Bandwidth reduction techniques
     • Application of federated machine learning in healthcare, Internet of Things (IoT), transportation system, tele-communications, and cybersecurity
2. Adversarial Machine Learning
   • What is it about?
     • Adversarial machine learning deals with fooling the machine learning model by supplying deceptive (often misleading or inaccurate) input, which may be in the form of gradients, dataset, etc. [2].
   • Why to use Adversarial Machine Learning?
     • With the growth of machine learning and introduction of decentralization systems, adversaries have been a major concern for machine learning techniques wherein they may introduce data violating the assumptions associated with the statistical distribution. In such cases, adversarial machine learning can be observed as a potential approach to enhance the robustness of the underlying model by generating attacks against the system for training it with regards to the adversarial attacks.
   • Active research areas:
     • Adversarial examples detection and it’s mitigation
     • Model inference attack and it’s mitigation
     • Robust optimization
     • Backdoor detection in decentralized ML environment
     • Creation of adversarial attacks
3. Natural Language Processing
   • What is it about?
     • Now-a-days, large amount of natural language-based data are generated in the form of news, speech, etc., and it is crucial to conduct the analysis of such textual data and extract knowledge. The goal of natural language processing is to make the computer capable of dealing with such analysis [3].
   • Active research areas:
     • Text categorization
     • Sentiment analysis
     • Text summarization
     • Text generation
     • Topic modelling
     • Fake news detection
4. Computer Vision
   • What is it about?
     • The multi-dimensional data we obtain these days from our surroundings are in the form of images, videos, symbols, etc. Computer vision deals with the acquisition, processing, and analysis of these images and videos, and extracting meaningful knowledge from that later assists in bringing automation in several image and video-related tasks (e.g., movement analysis, cell classification, etc.) [4].
   • Active research areas:
     • Pattern recognition
     • Image and video compression
     • Video analysis
     • Animation
     • Medical image analysis
     • Remote sensing
     • Adversarial computer vision
     • Sports analytics
     • Stochastic models
     • Self-less driving
     • Face recognition
     • Application of computer vision in military, healthcare, tactile feedback, and autonomous vehicles
5. Unsupervised Learning
   • What is it about?
     • Unsupervised learning is an active area of research to develop reliable algorithms for automated feature extraction and model training [5]. Now-a-days, untagged and unclassified data come in different formats and volumes, and researchers are concerned about automated extraction of features for model training. Unsupervised learning assists in learning patterns from such data to draw inferences.
   • Active research areas
     • Unsupervised algorithm development
     • Unsupervised learning in cybersecurity
     • Unsupervised learning in healthcare
6. Other topics for exploration
   • Few-shot learning [6]
   • Generative models [7]
   • Meta-learning [8]

Future Scope of Machine Learning and AI in 2022

Machine Learning (ML) and Artificial Intelligence (AI) have a bright future because they enable machines to gain information, which makes them more human-like. Machine learning is now being used in a wide variety of fields, maybe as many as one can envision. Read more

Related Article

• 12 Most Emerging Research Areas in Computer Science in 2021

References:

1. Yang, Q., Liu, Y., Chen, T., & Tong, Y. (2019). Federated machine learning: Concept and applications. ACM Transactions on Intelligent Systems and Technology (TIST), 10(2), 1-19.
2. Kurakin, A., Goodfellow, I., & Bengio, S. (2016). Adversarial machine learning at scale. arXiv preprint arXiv:1611.01236.
3. Chowdhury, G. G. (2003). Natural language processing. Annual review of information science and technology, 37(1), 51-89.
4. Forsyth, D., & Ponce, J. (2011). Computer vision: A modern approach (p. 792). Prentice hall.
5. Barlow, H. B. (1989). Unsupervised learning. Neural computation, 1(3), 295-311.
6. Sung, F., Yang, Y., Zhang, L., Xiang, T., Torr, P. H., & Hospedales, T. M. (2018). Learning to compare: Relation network for few-shot learning. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1199-1208).
7. Nalisnick, E., Matsukawa, A., Teh, Y. W., Gorur, D., & Lakshminarayanan, B. (2018). Do deep generative models know what they don’t know?. arXiv preprint arXiv:1810.09136.
8. Vilalta, R., & Drissi, Y. (2002). A perspective view and survey of meta-learning. Artificial intelligence review, 18(2), 77-95.

Cite this article:

K. Yadav, M. Quamara, B. Gupta (2021), 2021 Hot Topics in Machine Learning Research, Insights2Techinfo, pp.1

FAQ on this topic

73670cookie-check2021 Hot Topics in Machine Learning Researchyes