Abhijit Kulkarni

Leonard Sasse, Eliana Nicolaisen-Sobesky, Juergen Dukart et al.

Machine learning (ML) provides powerful tools for predictive modeling. ML's popularity stems from the promise of sample-level prediction with applications across a variety of fields from physics and marketing to healthcare. However, if not properly implemented and evaluated, ML pipelines may contain leakage typically resulting in overoptimistic performance estimates and failure to generalize to new data. This can have severe negative financial and societal implications. Our aim is to expand understanding associated with causes leading to leakage when designing, implementing, and evaluating ML pipelines. Illustrated by concrete examples, we provide a comprehensive overview and discussion of various types of leakage that may arise in ML pipelines.

Abhijit Kulkarni, Vinod John

Grid-interactive power converters are normally synchronized to the grid using phase-locked loops (PLLs). The performance of the PLLs is affected by the non-ideal conditions in the sensed grid voltage such as harmonics, frequency deviations and dc offsets in single-phase systems. In this paper, a single-phase PLL is presented to mitigate the effects of these non-idealities. This PLL is based on the popular second order generalized integrator (SOGI) structure. The SOGI structure is modified to eliminate of the effects of input dc offsets. The resulting SOGI structure has a high-pass filtering property. Hence, this PLL is termed as high-pass generalized integrator based PLL (HGI-PLL). It has fixed parameters which reduces the implementation complexity and aids in the implementation in low-end digital controllers. The HGI-PLL is shown to have least resource utilization among the SOGI based PLLs with dc cancelling capability. Systematic design methods are evolved leading to the design that limits the unit vector THD to within 1% for given non-ideal input conditions in terms of frequency deviation and harmonic distortion. The proposed designs achieve the fastest transient response. The performance of this PLL has been verified experimentally. The results are found to agree with the theoretical prediction.