GoSUMD.ai is a Software built by researchers at Aimdyn.com for complexity management. It offers the most comprehensive suite of design and analysis tools for engineers, consultants, modelers and data analysts.
Unlike the current fragmented techniques available for data analysis, GoSUMD.ai provides a closely integrated set of tools that unravel system complexity and allow users to see trends in data using easily interpreted graphical representations which were previously unattainable.
This provides deep insights based on a patented set of analysis techniques and visualization tools for complexity reduction, including global sensitivity analysis, uncertainty analysis, and model reduction.
GoSUMD.ai is based on some of the fastest available computational algorithms for nonlinear model representation and sampling, enabling simultaneous examination of literally tens of thousands of effects on the output of interest.
Due to fast computational algorithms, larger problems (size 100 – 10,000 parameters) can be examined with unprecedented accuracy.
How does it work?
GoSUMD.ai features a state-of the art sampling algorithm
Unlike the current fragmented techniques available for data analysis, GoSUMD.ai provides a closely integrated set of tools that unravel system complexity and allow users to see trends in data using easily interpreted graphical representations which were previously unattainable.
This provides deep insights based on a patented set of analysis techniques and visualization tools for complexity reduction, including global sensitivity analysis, uncertainty analysis, and model reduction.
GoSUMD.ai is based on some of the fastest available computational algorithms for nonlinear model representation and sampling, enabling simultaneous examination of literally tens of thousands of effects on the output of interest.
Due to fast computational algorithms, larger problems (size 100 – 10,000 parameters) can be examined with unprecedented accuracy.
How does it work?
GoSUMD.ai features a state-of the art sampling algorithm
- dynamical system based sampling algorithm (DSample) which provides an efficient way of uniformly sampling and features a better sampling performance than any other widely-adopted sampling method
- various modules where random sampling is inherently required are enhanced using DSample
- MATLAB models formatted as functions ( *.m files);
- Python models setup as scripts which take text files as input and produce textual output files.
- Type I models are learnt from data samples (input factors and corresponding outputs, e.g. real life data);
- Type II models are learnt from model executable and distribution type of input factors.
- Predict model outputs for a given set of input factors (predictors);
- Identify input factors responsible for model variability (global sensitivity);
- Perform model reduction;
- Quantify global uncertainty in model outputs.
- Optimize an objective function in the presence of equality and inequality constrains (both objective function and constrains can be functions of
- model outputs and model parameters);
- Perform global stability optimization.
- Fast handling of thousands of input factors;
- Automatic regularization of noisy/stochastic data/models;
- Fast handling of very large data sets for type I models;
- Optimal sampling of input factors' space (with optional constrains) for type II models;
- Handling both continuous and categorical input factors;
- Handling both continuous and multi-class outputs;
- Support of MATLAB models;
- Support of command-line usage of GoSUMD.ai via a configuration file.