Exclusive Lasso for Multi-task Feature Selection

By Yang Zhou, Rong Jin, Steven C. H. Hoi. 2010

228 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, background): “[40] consider a different scenario where useful features in different tasks have no overlapping.”
Latent Multi-Task Architecture Learning (2017, background): “• The (cid:96) 1 /(cid:96) 2 group lasso regularizer is =1 || G 1 ,i,g || 2 , a weighted sum over the (cid:96) 2 norms of the groups, often used to enforce subspace sharing (Zhou, Jin, and Hoi 2010; ´ Swirszcz and Lozano 2012).”
Sluice networks: Learning what to share between loosely related tasks (2017, background): “Group Lasso The (cid:96) 1 /(cid:96) 2 group lasso regularizer is =1 || G 1 ,i,g || 2 , a weighted sum over the (cid:96) 2 norms of the groups, often used to enforce subspace sharing [33, 26].”
Tree-Guided Group Lasso for Multi-Task Regression with Structured Sparsity (2009, background): “Compared to an arbitrary assignment of values for the group weights which can lead to an inconsistent estimate [5], the weights are systematically deﬁned in terms of the heights of the internal nodes in the tree, and each output variable is penalized in a balanced manner even if the groups overlap.”
Multi-task and Lifelong Learning of Kernels (2015, background): “These ideas were later generalized to the case, when related tasks may use slightly different kernel combinations [14,18], and successfully used in practical applications [15,16].”
Combined Group and Exclusive Sparsity for Deep Neural Networks (2017, methodology): “This exclusive feature learning is implemented by the exclusive sparsity regularization based on (1, 2)-norm (Zhou et al., 2010; Kong et al., 2014), which basically promotes network weights at each layer to compete for few meaningful features from the lower layer.”
Combinatorial Penalties: Structure preserved by convex relaxations (2017, methodology): “Sufficient support recovery and estimation consistency conditions for the (non-adaptive) estimator (5) have been established for homogeneous convex envelopes of submodular functions, for p =∞ in [1] and for general p in [29], in the high dimensional setting, and for latent group Lasso norm in [30], ”
On Multiplicative Multitask Feature Learning (2014, methodology): “These methods employ a regularizer based on the so-called (cid:96) 1 ,p matrix norm [12, 13, 15, 22, 24] that is the sum of the (cid:96) p norms of the rows in a matrix.”
Sparse Overlapping Sets Lasso for Multitask Learning and its Application to fMRI Analysis (2013, background): “Along related lines, the exclusive lasso [23] can be used when it is explicitly known that variables in certain sets are negatively correlated.”
Learning what to share between loosely related tasks (2017, background): “Group Lasso The (cid:96) 1 /(cid:96) 2 group lasso regularizer is =1 || G 1 ,i,g || 2 , a weighted sum over the (cid:96) 2 norms of the groups, often used to enforce subspace sharing (Zhou et al., 2010; ´Swirszcz and Lozano, 2012).”
Structured sparsity through convex optimization (2011, methodology): “A diﬀerent approach called “exclusive Lasso” consists instead of selecting a few variables inside each group, with some applications in multitask learning (Zhou, Jin and Hoi, 2010).”
Feature Selection Based on Structured Sparsity: A Comprehensive Study (2017, methodology): “[81] used negative correlation assumption through the l1,2-norm.”
Visual classification with multi-task joint sparse representation (2010, methodology): “We then naturally develop two kernel-view extensions of this method to deal with the situations where image descriptors and features are represented by kernel-based similarity matrices.”
Network of Experts for Large-Scale Image Categorization (2016, background): “where ||F ||e = √∑ j ( ∑ i Fij) 2 is the so-called exclusive lasso norm [5, 39],”