Current advanced hyperparameter optimization (HPO) methods, such as Bayesian optimization, have high sampling efficiency and facilitate replicability. Nonetheless, machine learning (ML) practitioners (e.g., engineers, scientists) mostly apply less advanced HPO methods, which can increase resource consumption during HPO or lead to underoptimized ML models. Therefore, we suspect that practitioners choose their HPO method to achieve different goals, such as decrease practitioner effort and target audience compliance. To develop HPO methods that align with such goals, the reasons why practitioners decide for specific HPO methods must be unveiled and thoroughly understood. Because qualitative research is most suitable to uncover such reasons and find potential explanations for them, we conducted semi-structured interviews to explain why practitioners choose different HPO methods. The interviews revealed six principal practitioner goals (e.g., increasing model comprehension), and eleven key factors that impact decisions for HPO methods (e.g., available computing resources). We deepen the understanding about why practitioners decide for different HPO methods and outline recommendations for improvements of HPO methods by aligning them with practitioner goals.

One way to reduce privacy risks for consumers when using the internet is to inform them better about the privacy practices they will encounter. Tailored privacy information provision could outperform the current practice where information system providers do not much more than posting unwieldy privacy notices. Paradoxically, this would require additional collection of data about consumers’ privacy preferences—which constitute themselves sensitive information so that sharing them may expose consumers to additional privacy risks. This chapter presents insights on how this paradoxical interplay can be outmaneuvered. We discuss different approaches for privacy preference elicitation, the data required, and how to best protect the sensitive data inevitably to be shared with technical privacy-preserving mechanisms. The key takeaway of this chapter is that we should put more thought into what we are building and using our systems for to allow for privacy through human-centered design instead of static, predefined solutions which do not meet consumer needs.

One barrier to more widespread adoption of differentially private neural networks is the entailed accuracy loss. To address this issue, the relationship between neural network architectures and model accuracy under differential privacy constraints needs to be better understood. As a first step, we test whether extant knowledge on architecture design also holds in the differentially private setting. Our findings show that it does not; architectures that perform well without differential privacy, do not necessarily do so with differential privacy. Consequently, extant knowledge on neural network architecture design cannot be seamlessly translated into the differential privacy context. Future research is required to better understand the relationship between neural network architectures and model accuracy to enable better architecture design choices under differential privacy constraints.To be presented at Privacy in Machine Learning (NeurIPS 2021 Workshop).

No abstract