Sökning: "First-order solution"

Visar resultat 1 - 5 av 76 avhandlingar innehållade orden First-order solution.

1. 1. First-Order Algorithms for Communication Efficient Distributed Learning

   Författare :Sarit Khirirat; Mikael Johansson; Tong Zhang; KTH; []
   Nyckelord :TEKNIK OCH TEKNOLOGIER; ENGINEERING AND TECHNOLOGY; Communication efficient learning; Optimization algorithms; Quantization; Error compensation; First-order algorithms; Stochastic gradient descent; Electrical Engineering; Elektro- och systemteknik;

   Sammanfattning : Innovations in numerical optimization, statistics and high performance computing have enabled tremendous advances in machine learning algorithms, fuelling applications from natural language processing to autonomous driving.To deal with increasing data volumes, and to keep the training times of increasingly complex machine learning models reasonable, modern optimization algorithms distribute both data and computations over a large number of machines. LÄS MER

3. 2. First-Order Algorithms for Communication Efficient Distributed Learning

   Författare :Sarit Khirirat; Mikael Johansson; Martin Jaggi; KTH; []
   Nyckelord :TEKNIK OCH TEKNOLOGIER; ENGINEERING AND TECHNOLOGY; Communication efficient learning; Optimization algorithms; Quantization; Error compensation; First-order algorithms; Stochastic gradient descent; Tillämpad matematik och beräkningsmatematik; Applied and Computational Mathematics; Optimeringslära och systemteori; Optimization and Systems Theory; Electrical Engineering; Elektro- och systemteknik;

   Sammanfattning : Technological developments in devices and storages have made large volumes of data collections more accessible than ever. This transformation leads to optimization problems with massive data in both volume and dimension. LÄS MER

4. 3. Hydrodynamic control of retention in heterogeneous aquifers and fractured rock

   Författare :Hua Cheng; Vladimir Cvetkovic; Chin-Fu Tsang; KTH; []
   Nyckelord :TEKNIK OCH TEKNOLOGIER; ENGINEERING AND TECHNOLOGY; Aquifer; Fracture; groundwater; First-order solution; Monte-Carlo simulation; Nu-clear waste; Heterogeneity; Retention processes; Matrix diffusion; Prediction; Evaluation; Travel time; Flow rate; ; Water engineering; Vattenteknik;

   Sammanfattning : In this thesis, fluid flow and solute transport in heterogeneous aquifers and particularly in frac-tured rock have been investigated using Lagrangian Stochastic Advective-Reaction (LaSAR) framework. The heterogeneity of the aquifer structure or fracture configuration, as well as the various reaction/retention processes have been considered in the modelling approach. LÄS MER

5. 4. Systems of Linear First Order Partial Differential Equations Admitting a Bilinear Multiplication of Solutions

   Författare :Jens Jonasson; Stefan Rauch; Sergio Benenti; Linköpings universitet; []
   Nyckelord :NATURVETENSKAP; NATURAL SCIENCES; Cauchy–Riemann equations; holomorphic functions; algebra; MATHEMATICS; MATEMATIK;

   Sammanfattning : The Cauchy–Riemann equations admit a bilinear multiplication of solutions, since the product of two holomorphic functions is again holomorphic. This multiplication plays the role of a nonlinear superposition principle for solutions, allowing for construction of new solutions from already known ones, and it leads to the exceptional property of the Cauchy–Riemann equations that all solutions can locally be built from power series of a single solution z = x + iy ∈ C. LÄS MER

7. 5. Convergence Analysis and Improvements for Projection Algorithms and Splitting Methods

   Författare :Mattias Fält; Institutionen för reglerteknik; []
   Nyckelord :TEKNIK OCH TEKNOLOGIER; ENGINEERING AND TECHNOLOGY; First-order methods; Convex Optimization; Nonsmooth Optimization; Large-scale Optimization; Iterative Methods; Douglas-Rachford splitting; Line Search;

   Sammanfattning : Non-smooth convex optimization problems occur in all fields of engineering. A common approach to solving this class of problems is proximal algorithms, or splitting methods. These first-order optimization algorithms are often simple, well suited to solve large-scale problems and have a low computational cost per iteration. LÄS MER