neos: End-to-End-Optimised Summary Statistics for High Energy Physics

The advent of deep learning has yielded powerful tools to automatically compute gradients computations. This is because training a neural network equates iteratively updating its parameters using gradient descent find the minimum loss function. Deep then subset broader paradigm; workflow with free that end-to-end optimisable, provided one can keep track all way through. work introduces neos: an example implementation following this paradigm fully differentiable high-energy physics workflow, capable optimising learnable summary statistic respect expected sensitivity analysis. Doing results in optimisation process aware modelling and treatment systematic uncertainties.