Complexity analysis of interior point algorithms for non-Lipschitz and nonconvex minimization
read more
Citations
Worst-case evaluation complexity for unconstrained nonlinear optimization using high-order regularized models
Implementable tensor methods in unconstrained convex optimization.
Iterative reweighted minimization methods for $$l_p$$lp regularized unconstrained nonlinear programming
Bilinear Factor Matrix Norm Minimization for Robust PCA: Algorithms and Applications
Interior Point Algorithms Theory And Analysis
References
Regression Shrinkage and Selection via the Lasso
The Elements of Statistical Learning: Data Mining, Inference, and Prediction
Introductory Lectures on Convex Optimization: A Basic Course
Regression shrinkage and selection via the lasso: a retrospective
Nearly unbiased variable selection under minimax concave penalty
Related Papers (5)
Variable Selection via Nonconcave Penalized Likelihood and its Oracle Properties
Cubic regularization of Newton method and its global performance
Frequently Asked Questions (12)
Q2. How can the second order interior point algorithm be used to solve a special case?
By using the Hessian of H, the second order interior point algorithm can generate an interior scaled second order stationary point in at most O( −3/2) steps.
Q3. How can the authors solve a convex optimization problem?
Using an interior-point algorithm, Ye [17] proved that an scaled KKT or first order stationary point of general quadratic programming can be computed in O( −1 log( −1)) iterations where each iteration would solve a ball-constrained or trust-region quadratic program that is equivalent to a simplex convex quadratic minimization problem.
Q4. What is the first order necessary condition for the first order interior point algorithm?
For the unconstrained l2-lp optimization (3), any local minimizer x satisfies the first order necessary condition [6]XAT (Ax− c) + λp|x|p = 0, (4)and the second order necessary conditionXATAX + λp(p− 1)|X|p 0, (5)where |X|p = diag(|x1|p, . . . , |xn|p).5
Q5. What is the first order interior point algorithm?
Lemma 3 If ρk ≥ 49 (γR 3 + 12λR p)‖dk‖ holds for all k ∈ K, then the Second Order Interior Point Algorithm produces an global minimizer of (1) in at most O( − 3 2 ) steps.
Q6. how many steps can be taken to compute the first term after the inequality?
The proposed Second Order Interior Point Algorithm obtains an scaled second order stationary point or global minimizer of (27) in no more than 36f(x0) max{λ2R2p, 8λRp} − 32 steps.
Q7. What is the solution to the SSQP problem?
(2)At each step, the SSQP algorithm solves a strongly convex quadratic minimization problem with a diagonal Hessian matrix, which has a simple closedform solution.
Q8. what is the first order interior point algorithm?
2.2 First Order Interior Point AlgorithmNote that for any x, x+ ∈ (0, b], Assumption 2.1 implies thatH(x+) ≤ H(x) + 〈∇H(x), x+ − x〉+ β 2 ‖x+ − x‖2. (7)Since ϕ is concave on [0,+∞), then for any s, t ∈ (0,+∞),ϕ(t) ≤ ϕ(s) + 〈∇ϕ(s), t− s〉.
Q9. What is the way to solve a first order stationary point?
The authors show that the objective function value f(xk) is monotonically decreasing along the sequence {xk} generated by the algorithm, and the worst-case complexity of the algorithm for generating an interior scaled first order stationary point of (1) is O( −2), which is the same in the worst-case complexity order of the steepest-descent methods for nonconvex smooth optimization problems.
Q10. What is the worst-case complexity of the second order interior point algorithm?
In Section 3, a second order interior point algorithm is given to solve a special case of (1), where H is twice continuously differentiable, ϕ := t and Ω = {x : x ≥ 0}.
Q11. What are the penalty functions in R++?
These six penalty functions are concave in R+ and continuously differentiable in R++, which are often used in statistics and sparse reconstruction.
Q12. what is the complexity of the second order interior point Algorithm?
∞In − λRp p(1− p)(2− p)(3− p) 2 ‖dk‖∞In − λ 2 Rp‖dk‖∞In.(58)From (54), (55), (56), (58), and ρk < 4 9 (γR 3 + λ2R p)‖dk‖, the authors obtainXk+1∇2f(xk+1)Xk+1 =Xk+1∇2H(xk+1)Xk+1 + λp(p− 1)Xk+1Xp−2k+1Xk+1− (9 4 ρk + γR 3‖dk‖∞ + 1 2 λRp‖dk‖∞)In − 2(γR3 + λ 2 Rp)‖dk‖∞In −min{1, 1 γR3 + 12λR p } √ In − √ In.(59)17According to Lemmas 3 - 5, the authors can obtain the complexity of the Second Order Interior Point Algorithm for finding an scaled second order stationary point of (27).