Towards a Reliable Statistical Oracle and its Applications
نویسنده
چکیده
It is shown how—based on the idea of the Heuristic Oracle—a Statistical Oracle can be implemented based on statistical tests. Whereas the decision of a Heuristic Oracle may be wrong, it will be demonstrated how this can be avoided with the Statistical Oracle, using techniques from the field of randomized algorithms. As with all types of oracles, the Statistical Oracle is not universially applicable. If explicit formulae for the mean, variance, or distribution of characteristics computable from the test output are available, it is possible to apply the Statistical Oracle. Especially in the field of image processing, where inputs can be very complex and are thus difficult to generate, random testing is very useful. It is shown, how the Statistical Oracle has been used to test implementations of image processing operations, namely dilation, erosion, and distance transform.
منابع مشابه
On Testing Image Processing Applications with Statistical Methods
Testing image processing applications is very ressource consuming. Many complex images have to be generated as test inputs and the expected resulting outputs have to be determined to complete the test cases. The present paper deals with this challenge in testing implementations of image operations, namely dilation. It applies random testing using models from stochastic geometry for random input...
متن کاملTowards Realizing Random Oracles: Hash Functions That Hide All Partial Information
The random oracle model is a very convenient setting for designing cryptographic protocols. In this idealized model all parties have access to a common, public random function, called a random oracle. Protocols in this model are often very simple and efficient; also the analysis is often clearer. However, we do not have a general mechanism for transforming protocols that are secure in the rando...
متن کاملA Framework for Quality Assurance of Machine Learning Applications
Some machine learning applications are intended to learn properties of data sets where the correct answers are not already known to human users. It is challenging to test and debug such ML software, because there is no reliable test oracle. We describe a framework and collection of tools aimed to assist with this problem. We present our findings from using the testing framework with three imple...
متن کاملSpotlight: Nafion-H®: Its catalytic applications
Fatemeh Karimi was born in 1984 in Shahreza/ Isfahan, Iran. She received her B.Sc. in pure Chemistry (2006) and M.Sc. in Organic Chemistry (2008) from Isfahan University of Technology, Iran, under the supervision of professor Abdol Reza Hajipour. She is currently working towards her Ph.D. under the supervision of professor Mohammad Ali Zolfigol at Bu-Ali Sina University. Her research interest i...
متن کاملAn Approach to Software Testing of Machine Learning Applications
Some machine learning applications are intended to learn properties of data sets where the correct answers are not already known to human users. It is challenging to test such ML software, because there is no reliable test oracle. We describe a software testing approach aimed at addressing this problem. We present our findings from testing implementations of two different ML ranking algorithms:...
متن کامل