Multi-task learning is becoming more and more popular. This post gives a general overview of the current state of multi-task learning. In particular, it provides context for current neural network-based methods by discussing the extensive multi-task learning literature.
1) Multi-class classification:
Multi-label classification (related problem of multi-output classification) is a task where an instance may be associated with multiple labels (labeling each sample with m labels from n_classes possible classes, where m can be 0 to n_classes inclusive).
There is no constraint on how many of the classes the instance can be assigned to. This can be thought of as predicting properties of a sample that are not mutually exclusive.
Ex: Prediction of the topics relevant to a text document or video.
The document or video may be about one of ‘religion’, ‘politics’, ‘finance’ or ‘education’, several of the topic classes or all of the topic classes.
Problem transformation:
Transformation into binary classification problems.
Transformation into multi-class classification problem.
Ensemble methods.
Adapted algorithms:
k-nearest neighbors
Decision trees
Kernel methods for vector output
Neural networks
References:
https://viblo.asia/p/multi-task-learning-mot-so-dieu-ban-nen-biet-3P0lPD08lox
https://colab.research.google.com/drive/1iErI9f8IGtXn1I3qiPbCmXTsJvrDmNQZ#scrollTo=KodLdjMTtOU7
https://towardsdatascience.com/multitask-learning-teach-your-ai-more-to-make-it-better-dde116c2cd40
https://pyimagesearch.com/2018/05/07/multi-label-classification-with-keras/
https://phamdinhkhanh.github.io/2020/04/22/MultitaskLearning.html
https://phamdinhkhanh.github.io/2020/05/05/MultitaskLearning_MultiBranch.html
https://pyimagesearch.com/2018/06/04/keras-multiple-outputs-and-multiple-losses/
