Roundtable 1: Rajesh K Parthasarathy

Bringing efficiency, consistency and assurance in creation of anonymized production datasets for analytical application development, testing and machine learning
How can you assure that all sensitive information in incoming source feeds is captured and validated for anonymization? How to maintain referential integrity with out losing functional quality in anonymized data sets?

Roundtable 2: Mark Cieliebak

How can we run a successful Artificial Intelligence project?
Established methods for software engineering projects are doomed to fail for AI/Machine Learning projects. How can we provide the customers early-on with reliable predictions of the results? But how can we efficiently aggregate and label this data? How can we guarantee to maintain a certain level of quality when the underlying data changes?

Roundtable 3: Manuel Renold

Deep Reinforcement Learning Doesn‘t Work Yet!?
How do you deal with the fact that small changes lead to nothing working anymore? Is Reinforcement Learning ready for industrial use? If so, who is dealing with? If not, what else is missing? The problem of overfitting, or the weird patterns in the environment.

Roundtable 4: Martin Epple

Visualization and communication of research results through interaction patterns
Data science often involves the process of visualizing unstructured and structured data in order to answer questions of social or business importance. Visualizing quantitative research results is associated with certain risks, e.g. the risk of visualizing disagreement on delicate issues. What are ways to mitigate this risk?

Time: 10:15 - 10:45
Track: Track 1