High-Dimensional Probability: An Introduction with Applications in Data Science
By Omiros Papaspiliopoulos. 2020
4,092 citations recorded by the source corpus.
© 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m ⪆ d ( T ) . For...
Subfields: Artificial Intelligence, Environmental Engineering, Statistics and Probability
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
- Yoshua Bengio
- Percy Liang
- Sergey Levine
- Jianfeng Gao
- Mohit Bansal
- Ruslan Salakhutdinov
- Quoc V. Le
- Michael I. Jordan
In Environmental Engineering
- Joel A. Tropp
- Wojciech Samek
- Per-Gunnar Martinsson
- Pushmeet Kohli
- Grgoire Montavon
- Balaji Lakshminarayanan
- H. Brendan McMahan
- Julyan Arbel
In Statistics and Probability
- Jayadev Acharya
- Jakub Konecný
- Per-Gunnar Martinsson
- Himanshu Tyagi
- Clément Louis Canonne
- Larry A. Wasserman
- Joel A. Tropp
- Peter Richtárik
Citations in the corpus
- Federated Learning with Autotuned Communication-Efficient Secure Aggregation (2019, background): “As a direct consequence of the isoperimetric inequality on the unit sphere [7], we find that”
- Benign overfitting in linear regression (2019, background): “1 from [37] implies that for some constant c1 and any i the random variable vzi is c1‖v‖(2)σ(2) x-subgaussian.”
- Confounding-Robust Policy Improvement (2018, methodology): “…of Lemma 3 with a contraction argument, and obtain a bound on the maxima of the absolute value deviation by an argument of Remark 8.1.5 of Vershynin (2018): note that the zero function is an element of the class of non-decreasing functions on R, and apply Dudley’s inequality to the…”
- Exploiting Shared Representations for Personalized Federated Learning (2021, methodology): “5 in Vershynin [2018] to bound the spectral norm of the d-by-k matrix in the RHS of (49).”
- Distributed Signal Detection under Communication Constraints (2020, positive): “(17), we finally obtain We now provide the last piece, the proof of Lemma 26. for some absolute constant C > 0 , using the fact that (cid:107) X ∼ χ 2 d is a subexponential random variable (see, e.g., (Vershynin, 2018, Section 2) (and in particular Proposition 2.7.1 and Exercise 2.7.10).”
- Quantitative stability of optimal transport maps under variations of the target measure (2021, neutral): “We say that a measure µ ∈ P (cid:16) R d (cid:17) is sub-exponential with variance proxy σ 2 for σ > 0 if it has zero mean and if for all r > 0, We refer to Proposition 2.7.1 in [35] for equivalent characterization.”
- Deep Generative Learning via Schrödinger Bridge (2021, background): “Since the diameter of NN φ under d NN is at most B , we have )) where c is a constant which may vary on different places, the first inequality follows from the chaining Theorem 8.1.3 in (Vershynin, 2018), the second inequality holds due to see (Bartlett et al., 2019).”
- Annealed Denoising Score Matching: Learning Energy-Based Models in High-Dimensional Spaces (2019, methodology): “To understand the learning process in relation to the data manifold X , we recall two elementary properties of random Gaussian vectors in high-dimensional spaces: First, the length distribution of random vectors becomes concentrated at √ dσ (Vershynin, 2018), where σ(2) is the variance of a single d”
- Prior Image-Constrained Reconstruction using Style-Based Generative Models (2021, neutral): “By Markov’s inequality and concentration of norm, we have the following (Vershynin, 2018):”
- Moving Beyond Sub-Gaussianity in High-Dimensional Statistics: Applications in Covariance Estimation and Linear Regression (2018, methodology): “This is one of the most commonly used type of tail assumptions on random vectors (especially with α = 2); see Section 3.4 of Vershynin (2018).”
- Neural Empirical Bayes (2019, background): “…an even deeper phenomenon, the concentration of measure, captured by a non-asymptotic result for the deviation of the norm X from √ d , where the deviation has a sub-gaussian tail (Tao, 2012; Vershynin, 2018): A related result states that In above expressions, C and c are absolute constants.”
- Minimax optimality of permutation tests (2020, methodology): “We then borrow the proof of Hanson–Wright inequality (see e.g. Rudelson and Vershynin, 2013; Vershynin, 2018) to proceed.”
- History Compression via Language Models in Reinforcement Learning (2022, background): “It can be found, e.g., in Vershynin (2018) lemma 5.3.2.”
- Uncertainty Principle for Communication Compression in Distributed and Federated Learning and the Search for an Optimal Compressor (2020, background): “For each I, we can find an ε-net for SδI in the Euclidean norm with cardinality at most (3/ε) δD (see Proposition 4.2.12 and Corollary 4.2.13 in (Vershynin, 2018)).”