Deep learning models have shown their strengths in various application domains. However, they often fail to comply with a given set of requirements defining the safe output space of the model. PiShield is the first framework ever allowing for the integration of the requirements into the neural networks' topology. Such integration happens in a straightforward and efficient manner and allows for the creation of deep learning models that are guaranteed to be compliant with the given requirements, no matter the input.
The requirements can be integrated both at inference and/or training time, depending on the practitioners' needs. Given the widespread application of deep learning models, there is a great need for frameworks allowing for the integration of the requirements in many different scenarios. We present three possible application domains: functional genomics, autonomous driving, and tabular data generation.