Estimation under group actions: Recovering orbits from invariants

被引:14
|
作者
Bandeira, Afonso S. [1 ]
Blum-Smith, Ben [2 ]
Kileel, Joe [3 ,4 ]
Niles-Weed, Jonathan [5 ,6 ]
Perry, Amelia [7 ]
Wein, Alexander S. [8 ]
机构
[1] Swiss Fed Inst Technol, Dept Math, Zurich, Switzerland
[2] Johns Hopkins Univ, Dept Appl Math & Stat, Baltimore, MD 21218 USA
[3] Univ Texas Austin, Dept Math, Austin, TX USA
[4] Univ Texas Austin, Oden Inst, Austin, TX USA
[5] NYU, Courant Inst Math Sci, New York, NY USA
[6] NYU, Ctr Data Sci, New York, NY USA
[7] MIT, Dept Math, Cambridge, MA USA
[8] Univ Calif Davis, Dept Math, Davis, CA USA
关键词
Signal processing; Biomedical imaging; Statistical aspects of information theory; Applications of lie groups; Applications of invariant theory; Applications theory Applications of commutative algebra; ZERO-SEPARATING INVARIANTS; CRYO-EM; POLYNOMIAL INVARIANTS; BISPECTRUM INVERSION; STRUCTURAL BIOLOGY; SAMPLE COMPLEXITY; DEGREE BOUNDS; RECONSTRUCTION; COMPUTATION; ALGORITHM;
D O I
10.1016/j.acha.2023.06.001
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
We study a class of orbit recovery problems in which we observe independent copies of an unknown element of Rp, each linearly acted upon by a random element of some group (such as Z/p or SO(3)) and then corrupted by additive Gaussian noise. We prove matching upper and lower bounds on the number of samples required to approximately recover the group orbit of this unknown element with high probability. These bounds, based on quantitative techniques in invariant theory, give a precise correspondence between the statistical difficulty of the estimation problem and algebraic properties of the group. Furthermore, we give computerassisted procedures to certify these properties that are computationally efficient in many cases of interest. The model is motivated by geometric problems in signal processing, computer vision, and structural biology, and applies to the reconstruction problem in cryo-electron microscopy (cryo-EM), a problem of significant practical interest. Our results allow us to verify (for a given problem size) that if cryo-EM images are corrupted by noise with variance & sigma;2, the number of images required to recover the molecule structure scales as & sigma;6. We match this bound with a novel (albeit computationally expensive) algorithm for ab initio reconstruction in cryo-EM, based on invariant features of degree at most 3. We further discuss how to recover multiple molecular structures from mixed (or heterogeneous) cryo-EM samples. & COPY; 2023 Elsevier Inc. All rights reserved.
引用
收藏
页码:236 / 319
页数:84
相关论文
共 16 条
  • [1] Reconstructing under group actions
    Radcliffe, A. J.
    Scott, A. D.
    GRAPHS AND COMBINATORICS, 2006, 22 (03) : 399 - 419
  • [2] Reconstructing under Group Actions
    A.J. Radcliffe
    A.D. Scott
    Graphs and Combinatorics, 2006, 22 : 399 - 419
  • [3] Rational Invariants of Even Ternary Forms Under the Orthogonal Group
    Goerlach, Paul
    Hubert, Evelyne
    Papadopoulo, Theo
    FOUNDATIONS OF COMPUTATIONAL MATHEMATICS, 2019, 19 (06) : 1315 - 1361
  • [4] Quantum Money from Abelian Group Actions
    Zhandry, Mark
    15TH INNOVATIONS IN THEORETICAL COMPUTER SCIENCE CONFERENCE, ITCS 2024, 2024,
  • [5] Group Regularized Estimation Under Structural Hierarchy
    She, Yiyuan
    Wang, Zhifeng
    Jiang, He
    JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 2018, 113 (521) : 445 - 454
  • [6] ANCESTRAL CHARACTER ESTIMATION UNDER THE THRESHOLD MODEL FROM QUANTITATIVE GENETICS
    Revell, Liam J.
    EVOLUTION, 2014, 68 (03) : 743 - 759
  • [7] Recovering task fMRI signals from highly under-sampled data with low-rank and temporal subspace constraints
    Chiew, Mark
    Graedel, Nadine N.
    Miller, Karla L.
    NEUROIMAGE, 2018, 174 : 97 - 110
  • [8] Multi-Exponential Transverse Relaxation Times Estimation From Magnetic Resonance Images Under Rician Noise and Spatial Regularization
    El Hajj, Christian
    Moussaoui, Said
    Collewet, Guylaine
    Musse, Maja
    IEEE TRANSACTIONS ON IMAGE PROCESSING, 2020, 29 : 6721 - 6733
  • [9] PRANC: ML species tree estimation from the ranked gene trees under coalescence
    Kim, Anastasiia
    Degnan, James H.
    BIOINFORMATICS, 2020, 36 (18) : 4819 - 4821
  • [10] Recovering fine details from under-resolved electron tomography data using higher order total variation l1 regularization
    Sanders, Toby
    Gelb, Anne
    Platte, Rodrigo B.
    Arslan, Ilke
    Landskron, Kai
    ULTRAMICROSCOPY, 2017, 174 : 97 - 105