Logo des Repositoriums

Auflistung nach Schlagwort "Regression"

1 - 2 von 2
  • Konferenzbeitrag
    Property-Driven Black-Box Testing of Numeric Functions
    (Software Engineering 2023, 2023) Sharma, Arnab; Melnikov, Vitalik; Hüllermeier, Eyke; Wehrheim, Heike
    In this work, we propose a property-driven testing mechanism to perform unit testing of functions performing numerical computations. Our approach, similar to the property-based testing technique, allows the tester to specify the requirements to check. Unlike property-based testing, the specification is then used to generate test cases in a targeted manner. Moreover, our approach works as a black-box testing tool, i.e. it does not require knowledge about the internals of the function under test. Therefore, besides on programmed numeric functions, we also apply our technique to machine-learned regression models. The experimental evaluation on a number of case studies shows the effectiveness of our testing approach.
  • Zeitschriftenartikel
    Remember to Correct the Bias When Using Deep Learning for Regression!
    (KI - Künstliche Intelligenz: Vol. 37, No. 1, 2023) Igel, Christian; Oehmcke, Stefan
    When training deep learning models for least-squares regression, we cannot expect that the training error residuals of the final model, selected after a fixed training time or based on performance on a hold-out data set, sum to zero. This can introduce a systematic error that accumulates if we are interested in the total aggregated performance over many data points (e.g., the sum of the residuals on previously unseen data). We suggest adjusting the bias of the machine learning model after training as a default post-processing step, which efficiently solves the problem. The severeness of the error accumulation and the effectiveness of the bias correction are demonstrated in exemplary experiments.