However, many use cases of neural network verification require solving the inverse problem, or over-approximating the set of inputs that lead to certain outputs.

In recent years, machine learning has shown great promise in a variety of real-world applications.

Deep Neural Networks (DNNs) have exceeded human capability on many computer vision tasks.

To bridge this crucial gap, first, this paper presents a comprehensive benchmark suite to evaluate the out-of-distribution (OOD) performance of image compression methods.

We study offline model-based optimization to maximize a black-box objective function with a static dataset of designs and scores.

Nonetheless, modern machine learning is adaptive in its nature. Prior information about a model's performance on the test set inevitably influences