| CPC G06F 16/215 (2019.01) [G06F 16/2365 (2019.01); G06F 16/285 (2019.01)] | 13 Claims |
|
1. A computer-implemented method of enabling an assessment of a plurality of datasets, each dataset of the plurality of datasets including a respective input datapoint in an input space and an associated output datapoint in an output space, the method comprising:
for each dataset of the plurality of datasets: determining multiple local complexity indicators for multiple neighborhoods of the input datapoint of the respective dataset in the input space, the multiple neighborhoods having different sizes, any given local complexity indicator being based on differences between input distances and output distances, the input distances being in the input space between the input datapoint of the respective dataset and the input datapoints of each of multiple further datasets in the respective neighborhood, the output distances being in the output space between the output datapoint of the respective dataset and the output datapoints of each of the multiple further datasets in the respective neighborhood;
determining an array data structure having at least a first array dimension and a second array dimension, the first array dimension resolving the multiple neighborhoods based on sizes thereof, the second array dimension resolving values of the local complexity indicators in a binned manner, wherein entries of the array data structure are indicative of a frequency of an occurrence of the values of the respective local complexity indicators at a respective size of the neighborhoods across all datasets of the plurality of datasets;
controlling a user interface to provide access to the array data structure;
controlling the user interface to receive a user selection of a subsection of the array data structure;
selecting a subset of the plurality of datasets associated with the subsection; and
controlling the user interface to output information associated with the subset of the plurality of datasets;
wherein the plurality of datasets define inference data and are obtained from an inference task provided by a machine-learning algorithm, and the method further comprises:
accessing the array data structure to assess the inference data;
controlling a technical system based on the inference task;
aborting the inference task in response to detecting a wrong prediction based on an assessment of the array data structure; and
when performing the aborting of the inference task, selectively transitioning the technical system to a safe state.
|