Visual Representations: Defining Properties and Deep Approximations
By Stefano Soatto, Alessandro Chiuso. 2014
66 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
- Invariant Representations without Adversarial Training (2018, neutral): “Models including “nuisance” factors were also considered by Soatto and Chiuso [19], in which the authors propose definitions of both nuisances and invariant representations.”
- Emergence of Invariance and Disentanglement in Deep Representations (2017, background): “Although Soatto (2013) addressed maximal invariants, we only consider sufficient invariants (Soatto and Chiuso, 2016).”
- Geometric Robustness of Deep Networks: Analysis and Improvement (2017, unknown): “The vulnerability of CNNs against geometric transformations, on the other hand, has been studied in [17] and [24] that analyze image and visual representations to find theoretical foundations of transformation invariant features.”
- What makes for good views for contrastive learning (2020, background): “This idea is related to the idea of minimal sufficient statistics [48] and the Information Bottleneck theory [54, 2], which have been previously articulated in the representation learning literature.”
- Information Dropout: learning optimal representations through noise (2017, background): “To address this problem, Soatto & Chiuso (2016) defined optimal representations not in terms of maximality, but in terms of sufficiency , and characterized representations that are both sufficient and invariant.”
- A Separation Principle for Control in the Age of Deep Learning (2017, methodology): “Per [36], if that approached the true posterior, it would be a minimal sufficient representation.”
- Generative Models of Visually Grounded Imagination (2017, background): “More generally, we argue that we need to have at least some labels or attributes in order to define what we mean by a good representation (Soatto and Chiuso, 2016).”
- Evading the Adversary in Invariant Representation (2018, background): “Models including confounds or “nuisance” factors were also considered by Soatto and Chiuso [18], in which the authors propose definitions of both nuisances and invariant representations.”
- Generalization Error of Invariant Classifiers (2016, background): “A prominent example is associated with the field of computer vision where the features and learning algorithms are designed to be invariant to the intrinsic variability in the data (Soatto and Chiuso, 2016).”
- Mathematics of Deep Learning (2017, background): “Indeed, even basic concepts such as sufficiency and invariance have received diverse treatment [9], [14], [15].”