We're pleased to introduce Version 2016-R2 of Frontline Solvers for Excel, including our flagship integrated product Analytic Solver Platform, and our professional entry-level integrated product Analytic Solver Pro. In recent years we've been offering "Christmas in July" -- a major release near year-end, and another major new release in mid-year -- but in 2016, July has moved up to April. This is our fourth new product release in 2016, following XLMiner.com, our new cloud analytics platform in January, XLMiner SDK in February, and Solver SDK V2016 in March.
New Optimization Power in Evolutionary Solver
V2016-R2 features significant enhancements to our Evolutionary Solver, focusing on models with integer variables – often yielding dramatically better solutions in a given amount of time, compared to our previous releases and competitive products. These enhancements build on the SQP-GS (Sequential Quadratic Programming with Gradient Sampling) for local search, and Feasibility Pump algorithms that we introduced four months ago in Frontline Solvers V2016. If you have a model for the Evolutionary Solver with integer variables, you're likely to see a real performance boost in V2016-R2 -- and we have more enhancements coming down the pike for models like yours.
New Support for ‘Sparse Cubes’ in Dimensional Modeling
This release also includes performance enhancements for multi-dimensional optimization and simulation models that use the Dimensional Modeling features of Analytic Solver Platform, in the frequently-occurring case where multi-dimensional data is ‘sparse’. Memory use and time computing multi-dimensional formulas is often dramatically reduced for these models. If you haven't explored Dimensional Modeling yet (see the Model dropdown on the Analytic Solver Platform Ribbon), you have some new reasons to do so -- check out the chapter "Dimensional Modeling" in the Frontline Solvers User Guide.
New Correlation Methods in Monte Carlo Simulation
In Monte Carlo simulation, V2016-R2 supports correlation of uncertain variables with dissimilar distributions, using copulas – Gaussian, Student and Archimedean (Clayton, Frank and Gumbel) forms – as an alternative to Risk Solver's existing support for rank-order correlation. Copulas for correlation are often used in quantitative finance models. All these correlation methods are available via the simple point-and-click Correlation dialog you may already know, allowing you to preview scatter plots showing how samples from uncertain variables will be generated and correlated.
New Support for Compound Probability Distributions
The new release also supports ‘compound distributions’ – often used in actuarial and insurance models – where samples are generated and aggregated from multiple instances of a probability distribution – called the ‘severity’ distribution. The number of instances can be a constant or a sample from another, discrete probability distribution, called the ‘frequency’ distribution. Nearly all of Risk Solver’s 50+ analytic distributions can be used, and for many common distributions, the software uses analytic methods in lieu of repeated sampling -- which means that you can still enjoy ultra-fast Monte Carlo simulation, even with compound distributions.
New Support for Tableau Web Data Connector
Except for the Dimensional Modeling performance enhancements which apply only to our Platform products, all these enhancements are available in Premium Solver Pro and Platform, Risk Solver Pro and Platform, and Analytic Solver Pro and Platform. We hope you'll enjoy and benefit from these latest product enhancements!