A convex formulation for learning shared structures from multiple tasks

By Jianhui Chen, Lei Tang, Jun Liu, Jieping Ye. 2009

210 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Mechanics, Optimization & Training

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 Computational Mechanics

In Optimization & Training

Citations in the corpus

An Overview of Multi-task Learning (2018, methodology): “This model is then generalized in [25] by adding a squared Frobenius regularization on W and this generalized model can be relaxed to have a convex objective function.”
A Survey on Multi-Task Learning (2017, methodology): “For example, methods in [11], [44], [45] use f(·) to restrict the trace of Θ as shown in problems (2) and (13).”
Modeling disease progression via multi-task learning (2013, unknown): “Alternatively, one can assume that the tasks share a common subspace (Ando and Zhang, 2005; Chen et al., 2009), or a common set of features (Argyriou et al.”
Learning Tree Structure in Multi-Task Learning (2015, unknown): “For example, low-rank structure based methods in [1, 3] can learn model parameters sharing a low dimensional subspace for multiple tasks, probabilistic MTL models [27, 25] place probabilistic priors on multiple tasks to learn the model parameters as well as identifying the task relations, regularize”
Safe Screening for Multi-Task Feature Learning with Multiple Data Matrices (2015, background): “…have shown that learning multiple related tasks (MTL) simultaneously often provides superior predictive performance relative to learning each tasks independently (Ando and Zhang, 2005, Argyriou et al., 2008, Bakker and Heskes, 2003, Evgeniou et al., 2005, Zhang et al., 2006, Chen et al., 2013).”
Multi-task Sparse Structure Learning with Gaussian Copula Models (2016, methodology): “Some methods assume a fixed structure a priori, while others try to estimate it from the data.”
Multiple/Single-View Human Action Recognition via Part-Induced Multitask Structural Learning (2015, background): “[7] demonstrated that the optimal solution of (13) can be derived by solving the following problem:”
Collaborating between Local and Global Learning for Distributed Online Multiple Tasks (2015, background): “Many real world problems include a number of related learning tasks, for which multi-task learning (MTL) [5, 9, 10, 29] methods are favorable, as they can learn multiple related tasks together so as to improve the performance of each task relative to learning them separately.”
Learning Multiple Tasks with a Sparse Matrix-Normal Penalty (2010, methodology): “Along th is direction, researchers proposed to infer task structure via principal components [1, 12], inde pendent components [30] and covariance [2, 3] in the parameter space, to select a common subset of fea tures [9, 23], as well as to use shared hidden nodes in neural networks [6, 11].”
Learning to Multitask (2018, background): “As revealed in the survey [34], there is another representative formulation for the feature-based approach [2, 3, 11] in multitask learning.”
Privacy-Preserving Distributed Multi-Task Learning with Asynchronous Updates (2017, methodology): “Examples of commonly used MTL regularizations include shared feature subspace via low-rank modeling [2, 11] and shared feature subset via group sparsity [12, 31, 46].”
Parallel Multi-task Learning (2015, methodology): “This model is further investigated in [6] by relaxing the nonconvex formulation in [5] to a convex one with a regularizer directly modeling the covariance among features.”
Learning Sparse Task Relations in Multi-Task Learning (2017, background): “By assuming that the model parameters of all the tasks share a low-rank subspace, Ando and Zhang (Ando and Zhang 2005) directly learn the shared subspace as well as the task-specific spanning coefficients under a non-convex formulation and then Chen et al. (Chen et al. 2009) relax the non-convex for”
Integrating low-rank and group-sparse structures for robust multi-task learning (2011, background): “One key assumption in most of the existing MTL algorithms is that all tasks are correlated via a certain structure, which, for example, includes hidden units in neural networks [7], a common prior in a hierarchical Bayesian model [3, 29, 35, 37], parameters in Gaussian process covariance [17], kerne”
Multi-Task Learning with Low Rank Attribute Embedding for Multi-Camera Person Re-Identification (2018, background): “The modeling of information shared across tasks is often based on the assumption of a shared low rank structure [53], [54].”