Concrete ML v1.5: Encrypted DataFrames and Faster Neural Networks

Concrete ML v1.5 introduces several enhancements, including a new DataFrame API that allows encrypted data manipulation, advancing its primary use case of private inference and confidential collaboration. This release also offers a feature that accelerates neural network inference by 2-3 times through a trade-off between prediction speed and precision, with an improved FHE simulation mode to assess the impact on accuracy. Furthermore, a new demo showcases secure text data anonymization for querying knowledge bases with ChatGPT without exposing personally identifiable information, enabling the creation of anonymized knowledge bases for retrieval-augmented generation. The DataFrame API draws inspiration from the Pandas package, providing capabilities for encrypting, joining, and decrypting data, facilitating encrypted collaboration and model training.