Gradient-based Optimization of the Area under the Minimum of False Positive and False Negative Functions

• Published in: 2021 2nd International Conference on Smart Electronics and Communication (ICOSEC) pp. 779 - 785
• Main Authors: Mrudula Devi, K, Venkata Ramakrishna, Sunkari, Rama Koteswara Rao, Ganga, Prasad, Chitturi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07-10-2021
• DOI: 10.1109/ICOSEC51865.2021.9591945

Collector Operating Characteristic (ROC) bends are plots of genuine awesome expense versus bogus eminent charge which are helpful for assessing twofold order models, but intense to use for becoming more acquainted with because of the way that the Area Under the Curve (AUC) is non-arched. ROC bends can moreover be utilized in various issues that have bogus awesome and real fine charges, for example, changepoint identification. This research work shows that in this noteworthy normal setting, the ROC bend is currently not typically monotonic, thus in case circles are existing the Area Under the Curve (AUC) can be higher than one. Also, this research work suggests another differentiable substitute misfortune include alluded to as the AUM, brief for Area Under Min (FP, FN). Finally, this research study shows that the AUM slope can be successfully figured and utilized in another dominating calculation. In our experimental look into of managed twofold arrangement and changepoint recognition issues, it has been shown that the proposed novelAUM minimization calculation results in expanded AUC and comparable speed comparative by going before baselines.