| 1. |
- Aytekin, Arda, 1986-
(författare)
-
Asynchronous Algorithms for Large-Scale Optimization : Analysis and Implementation
- 2017
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This thesis proposes and analyzes several first-order methods for convex optimization, designed for parallel implementation in shared and distributed memory architectures. The theoretical focus is on designing algorithms that can run asynchronously, allowing computing nodes to execute their tasks with stale information without jeopardizing convergence to the optimal solution.The first part of the thesis focuses on shared memory architectures. We propose and analyze a family of algorithms to solve an unconstrained, smooth optimization problem consisting of a large number of component functions. Specifically, we investigate the effect of information delay, inherent in asynchronous implementations, on the convergence properties of the incremental prox-gradient descent method. Contrary to related proposals in the literature, we establish delay-insensitive convergence results: the proposed algorithms converge under any bounded information delay, and their constant step-size can be selected independently of the delay bound.Then, we shift focus to solving constrained, possibly non-smooth, optimization problems in a distributed memory architecture. This time, we propose and analyze two important families of gradient descent algorithms: asynchronous mini-batching and incremental aggregated gradient descent. In particular, for asynchronous mini-batching, we show that, by suitably choosing the algorithm parameters, one can recover the best-known convergence rates established for delay-free implementations, and expect a near-linear speedup with the number of computing nodes. Similarly, for incremental aggregated gradient descent, we establish global linear convergence rates for any bounded information delay.Extensive simulations and actual implementations of the algorithms in different platforms on representative real-world problems validate our theoretical results.
|
|
| 2. |
- Herceg, Domagoj, et al.
(författare)
-
Data-driven Modelling, Learning and Stochastic Predictive Control for the Steel Industry
- 2017
-
Ingår i: 2017 25th Mediterranean Conference on Control and Automation, MED 2017. - Piscataway, NJ : Institute of Electrical and Electronics Engineers (IEEE). - 9781509045334 ; , s. 1361-1366
-
Konferensbidrag (refereegranskat)abstract
- The steel industry involves energy-intensive processessuch as combustion processes whose accurate modellingvia first principles is both challenging and unlikely to leadto accurate models let alone cast time-varying dynamics anddescribe the inevitable wear and tear. In this paper we addressthe main objective which is the reduction of energy consumptionand emissions along with the enhancement of the autonomy ofthe controlled process by online modelling and uncertaintyawarepredictive control. We propose a risk-sensitive modelselection procedure which makes use of the modern theoryof risk measures and obtain dynamical models using processdata from our experimental setting: a walking beam furnaceat Swerea MEFOS. We use a scenario-based model predictivecontroller to track given temperature references at the threeheating zones of the furnace and we train a classifier whichpredicts possible drops in the excess of Oxygen in each heatingzone below acceptable levels. This information is then used torecalibrate the controller in order to maintain a high qualityof combustion, therefore, higher thermal efficiency and loweremissions
|
|
