Curriculum learning of multiple tasks

By Anastasia Pentina, Viktoriia Sharmanska, Christoph H. Lampert. 2014

274 citations recorded by the source corpus.

Sharing information between multiple tasks enables algorithms to achieve good generalization performance even from small amounts of training data. However, in a realistic scenario of multi-task learning not all tasks are equally related to each other, hence it could be advantageous to transfer information only between the most related tasks. In this work we propose an approach that processes multiple tasks in a sequence with sharing between subsequent tasks instead of solving all tasks jointly. Subsequently, we address the question of curriculum learning of tasks, i.e. finding the best order of tasks to be learned. Our approach is based on a generalization bound criterion for choosing the task order that optimizes the average expected classification performance over all tasks. Our experimental results show that learning multiple related tasks sequentially can be more effective than learning them jointly, the order in which tasks are being solved affects the overall performance, and that our model is able to automatically discover a favourable order of tasks.

Subfields: Artificial Intelligence, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Reinforcement Learning

Citations in the corpus

Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges (2019, background): “We highlight a few open problems along these lines: Task Scheduling: Scheduling tasks has been widely studied in the context of multitask learning and meta-learning, but remains relatively under-explored for multilingual NMT (Bengio et al., 2009; Pentina et al., 2015).”
Cross-Stitch Networks for Multi-task Learning (2016, background): “and their widespread applications in other fields, such as genomics [38], natural language processing [7, 8, 35] and computer vision [3, 10, 30, 31, 40, 51, 53, 58].”
UberNet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory (2016, neutral): “[27, 46].”
iCaRL: Incremental Classifier and Representation Learning (2016, background): “[29] show that learning multiple tasks sequentially can beneficial, but for choosing the order the data for all tasks has to be available at the same time.”
Gradient Episodic Memory for Continual Learning (2017, background): “Curriculum learning considers learning a sequence of data [Bengio et al., 2009], or a sequence of tasks [Pentina et al., 2015], sorted by increasing difﬁculty.”
On Class Orderings for Incremental Learning (2020, background): “In contrast to curriculum learning, our objective is not to obtain the best model performance (Pentina et al., 2015), but to propose an evaluation for class-IL methods under different scenarios.”
Meta-Transfer Learning for Few-Shot Learning (2018, methodology): “[35] use adaptive SVM classifiers to evaluate task difficulty for later organization.”
End-to-End Relation Extraction using LSTMs on Sequences and Tree Structures (2016, methodology): “We follow the settings by (Li and Ji, 2014), and we did not use the full mention boundary unlike Lu and Roth (2015).”
Meta-Transfer Learning Through Hard Tasks (2019, methodology): “[63] use adaptive SVM classifiers to evaluate task difficulty for later organization.”
Dynamic Task Prioritization for Multitask Learning (2018, background): “Multitask learning benefits from multiple related tasks [63, 64] as they can reinforce one another and improve overall performance [23, 65].”
Continual Learning in Linear Classification on Separable Data (2023, unknown): “(e.g., in Li (2007); Pentina et al. (2015); Tang et al. (2022)).”
Growing Action Spaces (2019, methodology): “Many approaches use handcrafted schedules for task curricula, but others (Zaremba & Sutskever, 2014; Pentina et al., 2015; Graves et al., 2017) study diagnostics that can be used to automate the choice of task mixtures throughout training.”
A Survey on Curriculum Learning (2021, methodology): “CL then provides a sequence of task distributions to guide MTL, which is empirically proven helpful [29], [59], [83], [96].”
Adaptive SVM+: Learning with Privileged Information for Domain Adaptation (2017, background): “Curriculum learning [4, 22, 26], which is the learning strategy that implements this paradigm, employs the prior knowledge learned from the first “easier” tasks to improve the performance on “harder” ones that are learned at a later stage.”
FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling (2021, methodology): “This technique is widely employed in deep learning research [17–21].”
A Competence-aware Curriculum for Visual Concepts Learning via Question Answering (2020, background): “The competence-aware curriculum in our work is related to curriculum learning [7,53,55,20,51,44,21,47] and machine teaching [61,62,38,11,40,15,58].”
Brain-Inspired Model for Incremental Learning Using a Few Examples (2020, background): “Most of the earliest approaches did not fulfill the criteria for class-incremental learning and many required old class data to be available when learning new classes: [23, 33, 34, 32].”
Sentence-wise Smooth Regularization for Sequence to Sequence Learning (2018, background): “We also compare with a token-level curriculum learning (Bengio et al. 2009; Pentina, Sharmanska, and Lampert 2015) method.”