Continuous deep learning toolkit for real time adaptation II

Summary

The present document named “D5.4 Continuous deep learning toolkit for real time adaptation II.docx” is a public deliverable of the COMPOSITION project, co-funded by the European Union’s Horizon 2020 Framework Programme for Research and Innovation under Grant Agreement No 723145. It reports the final results of task “5.2 –Continuous Deep Learning Toolkit for real time adaptation” that foresees its development in work package 5 “Key Enabling Technologies for Intra-and Interfactory Interoperability and Data”.The document owner is ISMB and in this final version is a reiteration of the submitted at M16, namely D5.3 Continuous deep learning toolkit for real time adaptation I.This version highlightsthe final results after a second iteration of task 5.2, regarding the development of a Deep Learning Toolkit for real time adaptation. A comprehensive data assessment is provided alongside prove of concept results with an in-depth theoretical validation, for each of the addressed project’s use cases. Although this version has to be considered as final, there may still be afew software modifications since this deliverable is submitted at M30 whereas the project ends at M36.For the sake of completeness, this document adoptsa conservative approach in which valuable information from the previous iteration of this deliverable has been maintained in the annexes. Following this approach,the result isa solid document that can be redistributed among partners without lacking cross references and missing the opportunity to follow the logical evolution of tests, trials and deployments phases