Multilinear Multitask Learning

By Bernardino Romera-Paredes, H. Aung, Nadia Bianchi-Berthouze, Massimiliano Pontil. 2013

172 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.

Citations in the corpus

A Survey on Multi-Task Learning (2017, methodology): “Multilinear multi-task methods proposed in [115], [116] can handle this situation by”
Multi-task Deep Reinforcement Learning with PopArt (2018, background): “This is more challenging than the standard multi-task learning setup, which typically allows conditioning the model on the task index even at evaluation (Romera-Paredes et al. 2013; Collobert and Weston 2008), because our agents will need to infer what task to solve purely from the stream of raw obs”
An embarrassingly simple approach to zero-shot learning (2015, background): “For example if L is the sum of hinge losses, and Ω is the Frobenius norm, this would lead to a standard support vector machine (SVM), but one can consider other loss functions such as logistic loss, and other regularizers, such as the trace norm, leading to multitask learning methods [2, 32].”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, background): “On the other hand, the relation between DA and domain generalization [151, 254, 97, 94, 155], multi-task learning [26, 76, 178] or few-shot learning [149, 80] is much stronger.”
Deep Multi-task Representation Learning: A Tensor Factorisation Approach (2016, methodology): “Our proposed framework applies to this multiclass MDL setting where existing single-output MTL methods [Evgeniou and Pontil, 2004, Argyriou et al., 2008, Kumar and Daum´e III, 2012, Romera-paredes et al., 2013, Wimalawarne et al., 2014] do not.”
Learning to Transfer: Transferring Latent Task Structures and Its Application to Person-Specific Facial Action Unit Detection (2015, methodology): “We use the nonconvex formulation, reported to be the best in [18].”
Trace Norm Regularised Deep Multi-Task Learning (2016, background): “For example, sum of the trace norms on all matriciations [Romera-paredes et al., 2013], and scaled latent trace norm [Wimalawarne et al., 2014].”
Square Deal: Lower Bounds and Improved Relaxations for Tensor Recovery (2013, methodology): “In (Romera-Paredes et al., 2013; Xu et al., 2013), efficient methods based on alternating minimization are designed to solve related non-convex models.”
Multitask learning meets tensor factorization: task imputation via convex optimization (2014, methodology): “We consider supervised multitask learning problems [1, 6, 7] in which the tasks are indexed by a pair of indices known as multilinear multitask learning (MLMTL) [17, 19].”
Flexible Clustered Multi-Task Learning by Learning Representative Tasks (2016, methodology): “Among various MTL methods, clustered multi-task learning (CMTL) assumes that all tasks can be clustered into groups and attempts to learn the underlying cluster structure from the training data.”
Tensor Factorization for Low-Rank Tensor Completion (2018, background): “hyperspectral data recovery [6], [7], [12], image/video inpainting [1]–[3], text analysis [17], [18], and multitask learning [19], etc.”
A Unified Perspective on Multi-Domain and Multi-Task Learning (2014, methodology): “Some recent studies (Romera-paredes et al., 2013; Kishan Wimalawarne & Tomioka, 2014) noticed a drawback – this strategy can not represent a task with more structured metadata, e.”
Learning Multiple Tasks with Multilinear Relationship Networks (2015, methodology): “This fundamental idea of task relatedness has motivated a variety of methods, including multi-task feature learning that learns a shared feature representation [1, 2, 6, 5, 23], and multi-task relationship learning that models inherent task relationship [10, 14, 29, 31, 15, 17, 8].”
Continual Learning with Filter Atom Swapping (2022, methodology): “We are motivated by the literature on task subspace modeling (Evgeniou & Pontil, 2007; Maurer et al., 2013; Zhang & Yang, 2021; Romera-Paredes et al., 2013; Kumar & Daume III, 2012), where it is commonly assumed that task parameters lie in a low dimensional subspace, so that tasks can be modeled as…”
Multi-Task Feature Interaction Learning (2016, background): “In [27], Romera-Paredes et al. proposed a multilinear multitask (MLMTL) framework that arranges parameters of linear eﬀects from all tasks into a tensor W , by which they are able to represent the multi-modal relationships among tasks.”