Factorized Orthogonal Latent Spaces

By Mathieu Salzmann, Carl Henrik Ek, Raquel Urtasun, Trevor Darrell. 2010

128 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

Adversarial Continual Learning (2020, methodology): “es by learning merely the features that are specic to the task in hand and do not exist in z S. We further factorize z S and z P by using 6 S. Ebrahimi et al. orthogonality constraints introduced in [30], also known as \dierence" loss in the domain adaptation literature [4], to prevent the shar”
Deep Multimodal Feature Analysis for Action Recognition in RGB+D Videos (2016, methodology): “When features have very high dependency, descriptor-level fusion gives the best outcome, and when multiple groups of features have very low interdependency, kernel-level fusion performs better [51].”
Domain Separation Networks (2016, background): “Our main novelty is that, inspired by recent work [15, 25, 31] on shared–space component analysis, DSNs explicitly and jointly model both private and shared components of the domain representations.”
A Survey on Multi-view Learning (2013, methodology): “Salzmann et al. (2010) and Jia et al. (2010) proposed to find a latent subspace in which the information is correctly factorized into shared and private parts across different views.”
Latent Multi-Task Architecture Learning (2017, methodology): “task-specific) and shared space (Salzmann et al. 2010; Virtanen, Klami, and Kaski 2011) to more explicitly capture the difference between task-specific and cross-task features.”
Sluice networks: Learning what to share between loosely related tasks (2017, methodology): “Inspired by recent work on shared-space component analysis [31, 24], we add a penalty to enforce a division of labor and discourage redundancy between shared and task-speciﬁc subspaces.”
Bayesian group factor analysis with structured sparsity (2016, methodology): “Extensions of this approach to multiple coupled observations y (1) i ∈ R p 1 ×1 , · · · , y (m) i ∈ R pm×1 have resulted in group factor analysis models (GFA) (Archambeau and Bach, 2009; Salzmann et al., 2010; Jia et al., 2010; Virtanen et al., 2012).”
Adversarial Multi-task Learning for Text Classification (2017, background): “Motivated by recently work(Jia et al., 2010; Salzmann et al., 2010; Bousmalis et al., 2016) on shared-private latent space analysis, we introduce orthogonality constraints, which penalize redundant latent representations and encourages the shared and private extractors to encode different aspects…”
Multi-view Super Vector for Action Recognition (2014, methodology): “Partly inspired by the Gaussian mixture model (GMM) based Fisher Vector representation [23] and the Factorized Orthogonal Latent Spaces (FOLS) approach [25] for multiview learning, in this paper, we propose a Mixture model of Probabilistic Canonical Correlation Analyzers (M-PCCA), and utilize this m”
Learning cross-modality similarity for multinomial data (2011, background): “Learning shared latent spaces across modalities has been an active research topic in human pose estimation [17, 6, 16] and image domain transfers [15].”
Partial Multi-View Clustering (2014, background): “The former are extensions from single-view spectral clustering approaches with the help of some similarity measures (de Sa 2005; Zhou and Burges 2007; Kumar and Daumé 2011; Kumar, Rai, and Daumé 2011), whereas the latter generally try to identify a latent subspace across the multiple views (Hardoon ”
Unsupervised Domain Adaptation with Similarity Learning (2017, background): “While this is, in principle, a good idea (given the representations are trained to be indistinguishable), it leaves the shared representation vulnerable to contamination by noise that is correlated with the underlying shared distribution [47, 7].”
Learning Multiple Views with Orthogonal Denoising Autoencoders (2016, methodology): “We evaluated our approach with a toy dataset similar to [6], which can be generated in 10 lines of MATLAB code (listed in Algorithm 1).”
In Defense of the Learning Without Forgetting for Task Incremental Learning (2021, background): “This regularization contains two parts: an adversarial loss that makes the shared encoding task-independent [6] and a disentanglement loss that acts to remove the overlap between the private- and the sharedencoding [28].”
Non-Linear Domain Adaptation with Boosting (2013, methodology): “Shared latent variable models have also been proposed to learn a shared representation across multiple feature sources or tasks [4, 19, 6, 7, 21].”
MISA: Modality-Invariant and -Specific Representations for Multimodal Sentiment Analysis (2020, background): “The origins of shared-private [5] learning can be found in multiview component analysis [48].”
Learning what to share between loosely related tasks (2017, methodology): “Another line of work looks into separating the learned space into a private (i.e. task-speciﬁc) and shared space (Salzmann et al., 2010; Virtanen et al., 2011) to more explicitly capture the difference be-tween task-speciﬁc and cross-task features.”
Factorized Latent Spaces with Structured Sparsity (2010, background): “In particular, [20] proposed to encourage the shared-private factorization to be nonredundant while simultaneously discovering the dimensionality of the latent space.”
Deep Variational Canonical Correlation Analysis (2016, methodology): “For VCCA/VCCA-private, we use Gaussian observation models as the inputs are real-valued.”
Unsupervised Multi-Target Domain Adaptation: An Information Theoretic Approach (2018, background): “, independent between the domains) subspaces of the data [12].”