I-LAMM for sparse learning: Simultaneous control of algorithmic complexity and statistical error

Mathematics

• Algorithmic Complexity 100%
• Contraction 20%
• Convex Program 59%
• Eigenvalue 15%
• Estimator 14%
• Family 34%
• Framework 27%
• High-dimensional 19%
• Iteration 31%
• Learning 66%
• Linear Approximation 26%
• Local Approximation 29%
• Loss Function 24%
• Model 7%
• Nonconvex Optimization 28%
• Nonconvex Problems 28%
• Numerical Results 16%
• Optimality 20%
• Optimization Problem 17%
• Penalized Quasi-likelihood 37%
• Penalty Function 26%
• Performance 14%
• Phase Transition 24%
• Rate of Convergence 18%
• Tolerance 51%

Business & Economics

• Approximation 14%
• Contraction 19%
• Eigenvalues 22%
• Estimator 14%
• Guarantee 14%
• Loss Function 20%
• Optimality 22%
• Optimization Problem 16%
• Penalty Function 24%
• Performance 6%
• Phase Transition 25%
• Quasi-likelihood 26%
• Rate of Convergence 22%
• Tolerance 38%