SC²S Colloquium - May 18, 2016
|Date:||May 18, 2016|
|Time:||3:00 pm, s.t.|
Tobias Neuhauser: Simulation-Based Parameter Optimization for Pre-Crash Functions
There is a positive trend in the automotive industry towards an increasing number of driver assistance systems. Each of these systems are configured by a large set of parameters, whereas manual tuning of these parameters with respect to performance, robustness and among other things is a costly task. The objective within this work is the development of a method for a semiautomatic parameter optimization of pre-crash functions for driver assistance systems. The work deals with sensitivity analysis of parameters, modeling cost functions, optimization algorithms and investigates the application of metamodel-based optimization.
Karthikeya Sampa Subbarao: Large-scale elastic machine learning using sparse grid combination technique
Large scale machine learning (ML) often requires flexible specification of ML algorithms for dynamic scaling, depending on the availability of resources. In this work, we introduce a grid-based ensemble technique for regression, which provides the foundation for a framework capable of dynamic resource utilization and fault tolerance. We also introduce a technique based on gradient boosting for efficient selection of grids for refinement.