Performance assessment of data classifiers
AssigneeCerebrus Solutions Limited
InventorsDerek M Dempsey, Kate Butchart
Life of this patent
Estimating the term…
Abstract
A method for assessing the performance of a data classifier operable to generate an element of output data in response to an element of input data, such as a neural network, is disclosed. The method includes using the data classifier to generate elements of result output data in response to elements of test input data, determining a measure of difference between each element of test output data and each corresponding element of result output data, forming a distribution function of the measures of differences, and forming a measure of performance from the distribution function.
Claims22
A method of assessing the performance of a data classifier operable to generate an element of output data in response to an element of input data, the method comprising:
- providing test data comprising elements of test input data and corresponding elements of test output data;
- operating the data classifier to generate elements of result output data in response to the elements of test input data;
- determining a measure of difference between each element of test output data and each corresponding element of result output data;
- forming a distribution function of said measures of difference;
- and
- generating a measure of performance of the data classifier from said distribution function.
A method of assessing the performance of a data classifier operable to generate an element of output data in response to an element of input data, the method comprising:
- providing test data comprising elements of test input data and corresponding elements of test output data;
- operating the data classifier to generate elements of result output data in response to elements of the test input data;
- determining a measure of difference between each element of test output data and each corresponding element of result output data;
- applying a weighting function to said measures of difference to form weighted measures of difference;
- and
- generating a measure of performance of the data classifier from said weighted measures of difference.
A method of training a data classifier operable to generate output data in response to input data, the method comprising:
- training the data classifier;
- providing test data comprising elements of test input data and corresponding elements of test output data;
- operating the data classifier to generate elements of result output data in response to the test input data;
- determining a measure of difference between each element of test output data and each corresponding element of result output data;
- forming a distribution function of said measures of difference;
- generating a measure of performance of the data classifier from said distribution function;
- and
- optionally retraining the data classifier in response to said measure of performance.
A data classifier system comprising:
- a data classifier operable to generate elements of result output data in response to elements of test input data, said elements of test input data also corresponding to elements of test output data;
- a difference generator operable to determine a measure of difference between each element of test output data and each corresponding element of result output data;
- a distribution function generator operable to form a distribution function of said measures of difference;
- and
- a performance measure generator operable to generate a measure of performance of the data classifier from said distribution function.
Dependencies are inferred from the claim text.