This version of GeneXproTools can process an unlimited number of samples. In practice this means that you can create models using several dozens of thousands samples thus increasing the level of trust of the models by exposing them to a large set of cases. This is valid for all editions of GeneXproTools and applies to training data, testing data and scoring data. GeneXproTools can easily handle a few million cells of data with very good data load times and a compact file format.

The new Enterprise Edition supports an unlimited number of variables. This is especially useful for the classification of DNA Microarrays, for example. GeneXproTools produces excellent models in a few seconds or minutes using sets well in excess of 25,000 variables on any modern desktop computer. This new capability also provides very interesting possibilities in the realms of data filtering and feature selection.

The Standard Edition of GeneXproTools 4.0 was expanded from 5 independent variables to 16, whereas the Advanced Edition was expanded from 20 to 64 independent variables. Furthermore, a new Academic Enterprise supporting unlimited variables is also available.

GeneXproTools features five different frameworks covering Function Finding, Logistic Regression, Classification, Time Series Prediction, and Logic Synthesis. Each framework is optimized to its problem domain and has specific data visualization, functions, fitness functions and code generation facilities. However, at their core, they all bring together the power of Gene Expression Programming and Statistics.

The models created by GeneXproTools generalize outstandingly well and as long as a representative set of training samples is used there is no risk of the model overfitting the data and the learning process can continue indefinitely with better and better models being created with time. Notwithstanding, GeneXproTools allows the use of a testing set for measuring the generalization capabilities of the evolved models and also performs an extensive statistical analysis on the testing data.

GeneXproTools supports custom fitness functions written in Javascript that you can easily add to any run. The custom fitness function can load external components and delegate the processing or load external datasets for complex fitness function. The fitness function has access to the relevant model information including model size, the model results and all the target data.

This very large spectrum of fitness functions allows you to judge your models using the measures you are used to work with. Most standard statistical measures are covered and for each of them there is a counterpart with parsimony pressure. The new functions also allow the exploration of different fitness landscapes increasing the chance of finding an excellent model for your problem.

The largest pool of operations in the industry is now accessible to your models including a section of decision rules with three and four arguments. The new Logic Synthesis algorithm includes dozens of rules and all the combinations for four variable ULM’s. The User Defined Functions and Dynamically Defined Functions also benefited from the fast script environment with very good performance. The Functions panel includes random function selection, function search and blanket weight changes.

The Function Selection facility of GeneXproTools 4.0 allows you to experiment with different function sets very quickly either by pressing the Random button or by selecting certain sub-sets of functions.

A brand new algorithm that dynamically adjusts the number of functions to the data environment compensating for very wide datasets. Coupled with finer adjustments at the function level through individual function weighting this feature simplifies the selection of appropriate function sets and provides better defaults for generic model creation.

Dynamic User Defined Functions (DDFs) behave exactly like the built-in functions of GeneXproTools and therefore can be used in the same manner to evolve models with them. On the other hand, Static User Defined Functions (UDFs) are useful for expressing known relationships between certain variables in datasets as GeneXproTools can then use them to create more complex models composed of several simpler building blocks.

All the best-of-generation models discovered during a run are saved and you can pick them up for analysis at the History Panel by checking the model you are interested in. Each of these intermediate models can then be analyzed exactly as you do for the best-of-run model, that is, you can check their performance in the testing set, check their vital statistics, automatically generate code with them, visualize their parse trees and use them as seed to create better models from them.

With GeneXproTools 4.0 you can use an existing model (which could have been either generated by GeneXproTools or by another modeling tool) as the starting point of an evolutionary process so that more complex, finely tuned models could be created. For models created outside of GeneXproTools or for GeneXproTools models modified by the user, the starting model or seed is fed to the algorithm through the Change Seed window where both the fitness and structural soundness of the model are tested. For models created in the GeneXproTools 4.0 environment, the seed (the best or active model of the previous run) is fed to the algorithm every time you click the Optimize or Simplify buttons in the Run Panel.

Optimization mode that reduces the size of a model while improving its accuracy at the same time. The Simplify Method reduces the redundancy in the model to an absolute minimum creating very lean models that are easier to study and less computer intensive in production environments. When used with Logic Synthesis it reduces the number of gates needed for the logic circuit, reducing costs and memory footprint.

When a model goes out of date and needs a little refreshing, you just have to replace the old training set (that is, the data set that was used to create the model you want to refresh) with the new one and then click Optimize in the Run Panel.

GeneXproTools supports most types of databases through ODBC and includes enhanced support for Microsoft SQL Server and Oracle, Access database and Excel files. For the very large datasets GeneXproTools uses a high performance text file manipulation algorithm that supports fast loading and several data transformations.

In the Data Panel you can visualize the distribution of all the inputs and spot simple relationships between the independent variables and the dependent variable. Curve fitting and Target/Model comparison plots in the Run Panel allow you to monitor how well the evolving models are fitting the target. In the Results/Predictions Panel you can compare the output of your model with the target output not only on the entire training set but also on the entire testing set using different charts.

Curve fitting charts in two and three dimensions are available for Function Finding and Time Series Prediction and target/model comparison charts for Classification and Logic Synthesis. Extensive statistical analyses (R-square, correlation coefficient, mean squared error, root mean squared error, mean absolute error, relative squared error, root relative squared error, relative absolute error, classification error, classification accuracy, confusion matrix, true positives, true negatives, false positives, false negatives, sensitivity, specificity, positive predictive value, and negative predictive value) are performed both on the training and testing sets for immediate evaluation of the predictive accuracy of the model in real use. In addition, you can also replace the testing set for further checking model generalization and manually tweak with the model itself.

The models can be translated into as many as sixteen programming languages (Ada, C, C++, C#, Fortran, Java, Java Script, Matlab, Pascal, Perl, PHP, Python, Visual Basic, VB.Net, Verilog, and VHDL) including specialty languages like VHDL and Verilog and languages relevant to the bio-informatics field like Perl and Python and the simulation fields like Ada and Matlab.

With your own grammars you can translate automatically the models evolved by GeneXproTools if you happen to prefer a programming language not already covered by the 16 built-in grammars of GeneXproTools 4.0 (Ada, C, C++, C#, Fortran, Java, Javascript, Matlab, Pascal, Perl, PHP, Python, Visual Basic, VB.Net, Verilog, and VHDL).

With the parse tree generator of GeneXproTools you can automatically convert your models into diagram representations or parse trees for a quicker and more complete understanding of their mathematical/logical intricacies.

The charts in the run panel provide invaluable information about the model and an insight into the algorithm inner workings. The charts show what variables are being used and which have a heavier weight, the model’s size and its tendency over time. Also exciting are the curve-fitting and target/model comparison charts that show how the evolving models are fitting the data. In addition, charts with the size of each sub-program and the total program size and charts with the sizes and fitness of all the chromosomes are also available for a better understanding of the evolutionary process.

With the scoring engine of GeneXproTools you can score your databases/Excel or text files immediately. The scoring engine of GeneXproTools uses the Javascript code of your models to score your data. The scores are saved to Excel or text files and can also be visualized in the Scoring Panel.

The default run templates of GeneXproTools allow you to start the modeling process with just a mouse click and yet obtain excellent results. They also make the discovery of good models quite enjoyable and fast as small populations of evolving models were chosen for all of them.

A new and fresh look with excellent access to all the areas of GeneXproTools and many navigation and organization options including a fully customizable toolbar. The wizards are short and simple to use and provide guidance throughout.