./nugetz

#numerical

43 packages tagged with “numerical

DiffSharp

DiffSharp is an automatic differentiation (AD) library. AD allows exact and efficient calculation of derivatives, by systematically invoking the chain rule of calculus at the elementary operator level during program execution. AD is different from numerical differentiation, which is prone to truncation and round-off errors, and symbolic differentiation, which is affected by expression swell and cannot fully handle algorithmic control flow. Using the DiffSharp library, derivative calculations (gradients, Hessians, Jacobians, directional derivatives, and matrix-free Hessian- and Jacobian-vector products) can be incorporated with minimal change into existing algorithms. Diffsharp supports nested forward and reverse AD up to any level, meaning that you can compute exact higher-order derivatives or differentiate functions that are internally making use of differentiation. Please see the API Overview page for a list of available operations. The library is under active development by Atılım Güneş Baydin and Barak A. Pearlmutter mainly for research applications in machine learning, as part of their work at the Brain and Computation Lab, Hamilton Institute, National University of Ireland Maynooth. DiffSharp is implemented in the F# language and can be used from C# and the other languages running on Mono or the .Net Framework, targeting the 64 bit platform. It is tested on Linux and Windows. We are working on interfaces/ports to other languages.

v0.7.756.2K
DifferentiationAutomaticSymbolicNumericalOptimization
pkg

OOOT

A library (.dll) of various linear, nonlinear, and stochastic numerical optimization techniques. While some of these are older than 40 years, many have yet to take advantage of an object-oriented programming model. This toolbox was originally created to aid in the automated design of various engineering artifacts. The following “class diagram” shows the main optimization methods that have currently been implemented – as well as some insight into organization of classes and objects. The best way to understand how to use OOOT, is through examples. There are two basic types of methods implemented so far: methods for continuous variables and methods for discrete variables. Approaches for mixed methods such as Branch-and-Bound are planned but not currently included. ClassDiagram1Many of the optimization methods for real valued decision variables will require that a line-search method be specified (Arithmetic Mean, Golden Section, DSC-Powell) and a search direction method (Cyclic Coordinate Search, Steepest Descent, Powell’s Method, Fletcher-Reeves, or Broyden-Fletcher-Goldfarb-Shanno). With this modular design it is possible to reconfigure these to create 80 unique algorithms. In addition to these there are also ten different convergence criteria that can be added individually or in any combination to these 80 algorithms. Read more about convergence methods here. But, one of the biggest advantages of the toolbox is the way that objective functions, equalities and inequalities can be easily added to the optimization process. Most (if not all open source approaches) optimization codes requires the problem specific details to be compiled along with the algorithm. Here, the problem formulation can be in another library or exe and can be customized at run-time instead of compile time. It is even possible to have the optimization algorithm change in response to the nature of the design variables and the number of constraints. Read more about problem formulation

v1.0.21.42222.9K
numericaloptimizationnonlinearprogramming