TestINT
TestINT is a test platform developed to be used in the solution of various problems brought by machine learning software that has a very wide application area today. TestINT includes artificial intelligence functions for intelligent systems and is also used in safety critical systems.
TESTINT, helps companies to develop products (with higher realibility) that have large learning data (and higher realibility.)
It allows the selection of different test techniques with its adaptable, flexible and expandable nature.
- TestINT is a test system developed for the following purposes for the effective and measurable testing of software developed with machine learning:
- Providing methods for the creation of a large sums of test data that allow testing different and extreme scenarios such as normal and heavy snow, desert conditions, night, stormy weather conditions,,
- Adaptation and further development of methods to be selected from traditional test methods for machine learning field
- Establishing criteria and making measurements to determine test adequacy,
- It provides standardization to supervise the reliability of software
Main Usage Purposes of TestINT
1. Preparetion of test data
Synthetic data generation with Generative Adversarial Networks (GAN)
-Image Proccessing and Transformation
-Image Proccessing and Transformation
-Creating Adversarial test data
-Data generation with domain specific constraints
-Image Proccessing and Transformation
-Image Proccessing and Transformation
-Creating Adversarial test data
-Data generation with domain specific constraints
2. Determination of Test Qualification Criteria
One layer neuron coverage
Basic layer neuron coverage
K-multisction neuron coverage
Neuron border coverage
Qualifiaciton (Satisfaction) based on test data
Realibility of k-nearest neighbors algorithm
Autoencoder
3. Test Run and Reporting Results
Basic layer neuron coverage
K-multisction neuron coverage
Neuron border coverage
Qualifiaciton (Satisfaction) based on test data
Realibility of k-nearest neighbors algorithm
Autoencoder
3. Test Run and Reporting Results
Test data augmentation using selected methods
Tests are run according to the selected test qualification criteria
The results are presented to the user with a test report
With new generated test data it allows access to faulty data
Tests are run according to the selected test qualification criteria
The results are presented to the user with a test report
With new generated test data it allows access to faulty data