Online Learning of Multiple Tasks and Their Relationships

By Avishek Saha, Piyush Rai, Hal Daumé Iii, Suresh Venkatasubramanian. 2011

132 citations recorded by the source corpus.

How researchers see this paper

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

In Reinforcement Learning

Citations in the corpus

An Overview of Multi-task Learning (2018, methodology): “When the training data in multiple tasks come in a sequentialway,traditionalMTLmodelscannothan-dlethembutmulti-taskonlinelearningiscapable ofdoingthisjob,asshowninsomerepresentative works[74–79].”
A Survey on Multi-Task Learning (2017, methodology): “In [157], an online version of the MTRL method is proposed to update both the model parameters and task covariance in a sequential way.”
Lifelong Learning for Sentiment Classification (2015, background): “Multi-task learning optimizes the learning of multiple related tasks at the same time (Caruana, 1997, Chen et al., 2011, Saha et al., 2011, Zhang et al., 2008).”
ELLA: An Efficient Lifelong Learning Algorithm (2013, methodology): “We report results using the LogDet update rule for OMTL, and found that the results did not vary greatly when other rules were employed; see (Saha et al., 2011) for more details on OMTL.”
Online Learning: A Comprehensive Survey (2018, background): “Some later approaches [146, 147] try to learn the rela-”
Active Lifelong Learning With "Watchdog" (2018, background): “It has also been shown to generally reduce the size of training data (Saha et al. 2011; Zhang 2010).”
Adaptive Smoothed Online Multi-Task Learning (2016, methodology): “Next, we implemented two online multi-task learning related to our approach: FOML – initializes η with fixed weights [8], Online Multi-Task Relationship Learning (OMTRL) [9] – learns a task covariance matrix along with task parameters.”
Robust Online Multi-Task Learning with Correlative and Personalized Structures (2017, methodology): “Instead of fixing the task relationship via a presumed structure [12], a recent Online Multi-Task Learning approach introduced an adaptive interaction matrix which quantified the task relevance with LogDet Divergence (OMTLLOG) and von-Neumann Divergence (OMTLVON) [7], respectively.”
Online Multitask Relative Similarity Learning (2017, unknown): “In [Saha et al. , 2011], Saha et al. proposed to adaptively update the relationship matrix via an incremental method.”
Confidence Weighted Multitask Learning (2019, methodology): “Our baselines cover three categories of MTL techniques: 1) two batch learning techniques: multitask feature learning ( MTFL ) (Argyriou, Evgeniou, and Pontil 2006) and trace-norm regularized MTL ( TRML ) (Zhou, Chen, and Ye 2011); 2) three online learning algorithms: online MTL ( OMTL ) (Saha et al.”
Smart Multitask Bregman Clustering and Multitask Kernel Clustering (2015, methodology): “The other way is to learn the intertask covariance matrix of the multivariate Gaussian prior which can model both positive and negative task correlations [Fei and Huan 2013; Saha et al. 2011; Zhang and Yeung 2010, 2012a, 2012b, 2014].”
Asymmetric multi-task learning based on task relatedness and loss (2016, methodology): “Saha et al. (2011) present an online multi-task learning method that is relevant to our work, where the pairwise relationships between tasks are modeled using a positive definite matrix.”
Multi-task Learning with Labeled and Unlabeled Tasks (2016, background): “by learning the relatedness of tasks (Saha et al., 2011; Kang et al., 2011) or sharing only between subgroups of tasks (Xue et al.”
Sentiment Classification by Leveraging the Shared Knowledge from a Sequence of Domains (2019, background): “LL is also different from multitask learning [3], which tries to optimize the learning of multiple tasks simultaneously [17,30,45].”
Using Task Descriptions in Lifelong Machine Learning for Improved Performance and Zero-Shot Transfer (2017, methodology): “Saha et al. (2011) use a task interaction matrix to model task relations online and Dekel et al. (2006) propose a shared global loss function that can be minimized as tasks arrive.”
Multi-Domain Learning: When Do Domains Matter? (2012, methodology): “Saha et al. (2011) improve upon the idea of simply using a ﬁxed task relationship matrix by instead learning it adaptively.”