Concrete v2.6: Approximate PBS, Input Compression, Extended Composable Functions, Speed Improvements

Concrete v2.6 introduces several advancements to enhance the performance, efficiency, and functionality of its framework, focusing on approximate programmable bootstrapping (PBS), input compression, extended function composability, and faster simulations. Approximate PBS involves improved rounding techniques that increase the speed of non-linear operations, beneficial for applications tolerant to inexact calculations like deep learning tasks. Input compression is further optimized to significantly reduce bandwidth and disk space usage, with input ciphertexts becoming much smaller. The update also advances function composability, allowing developers to define a set of functions within modules, enabling more flexible programming over encrypted data. The simulation speed has been improved dramatically, making development more efficient without requiring changes to the existing API. Enhancements in operators such as ternary-if, Relu, and Sign functions lead to faster processing, contributing to the overall efficiency of the platform. These improvements are designed to be seamlessly integrated into current workflows, encouraging users to explore the new features without needing to modify their existing setups.