WP1: Management

Provided efficient management and administration of the project, while fulfilling all legal requirements stated in EC rule and regulations and in the Consortium Agreement.
Established and maintained a functional management structure to ensure efficient communication between the partners.
Provided assistance and advice to the partners regarding administration and reporting.
Intellectual Property (IP) management.
Risk management and contingency planning.
Identified potential problems at an early stage and provided timely and effective solutions.
Developed and implemented a review and assessment structure to monitor project results with respect to objectives.

WP2: AIDApy: AI software framework, performance and new algorithms

Evaluated the available AI software frameworks on multiple architectures and selected the most robust option.
Improved the computing performances of the selected software framework.
Analyzed the underlying mathematical properties of the AI algorithms implemented.
Explored new modern algorithms for the training and deployment of AI algorithms for the analysis of Big Datasets.

Developed a high-level python front-end software that interfaces to open-source machine learning libraries. The front-end has several degrees of verbosity, allowing for both a non-expert and an expert exploitation of the machine learning libraries.
Analyzed satellite data with unsupervised machine learning algorithms for the following purposes: discovering latent variables through (nonlinear) dimensionality reduction; classifying “lookalike” events via clustering algorithms; infer causality relationship between (latent) variables through information theoretical tools.
Trained machine learning algorithms for supervised classification based on labeled data (either from real and synthetic data), that would be able to identify certain kind of space weather events.
Used the classification algorithm to create a catalog of interesting events.

Developed an open source code for the statistical analysis of variables, dedicated to in situ space missions.
Unified analysis methods developed by individual researchers, providing a complete suite of data analysis tools for heliospheric experts. This goal was achieved interconnecting the algorithms within the python language, starting from FORTRAN and C++ libraries.
Collected data from the AIDAdb, interacting with all the work-packages.
Extracted important information about turbulence, extreme/rare phenomena and hazardous events in the heliosphere, developing also new analysis tools for the AIDApy.
Increased performances via parallel computing.
Developed a user-friendly platform and a GUI wizard.

Developed python softwares able to provide measurements of fields and particles by means of virtual spacecraft flying through the output data of multi-dimensional simulations of any type (single as well as multi spacecraft configuration will be implemented), collected in AIDAdb.
Developed interpolation softwares to design virtual spacecraft trajectories within multi-dimensional simulation box; for the output of global simulations, real spacecraft trajectories will be implemented.
Developed softwares to feed with 3D particle Velocity Distributions (VDs) from numerical simulations, in order to simulate the response of a real particle instrument (virtual top-hat).
Developed techniques to estimate the local enhancement of plasma collisionality along a virtual spacecraft time series, due to generation of non-thermal features in the particle VDs.
Comparatively analyzed the output of the simulations and the measurements from virtual and real spacecraft (from AIDAdb), in order to support the design and training phase of novel techniques of artificial intelligence (AI) for data analysis and interpretation (AIDApy), aiming at identifying regions of scientific interest, based on time series of fields and particle VDs.

Enhanced an existing code for global magnetospheric simulations, OpenGGCM, with Data Assimilation techniques to improve its adherence to reality.
Produced DA-enhanced simulations of specific events selected in collaboration with WP8.
Fed simulation results into WP7 for inclusion into the Low Level and High Level Data Base, where they were used for training of ML techniques and signature recognition.

Defined and maintain the Data Management Plan (DMP) and AIDAdb.
Selected numerical codes and simulations used to produce data for ML training.
Compared event lists generated by other WP with automatic and non automatic techniques.
Gave user feedback on AIDApy.

Developed a python tool to automatically select and download data from a variety of open-access in situ, remote and ground plasma data archives.
Developed a python tool to process data and create lists of events by using routines based on human experience-driven data selection for different physical processes of interest (magnetic reconnection, turbulence, particle acceleration, etc.).
Integrated both tools into an open-access python software and interfaced this tool with AIDA databases and external databases

Optimized, package and delivered the ML software engine developed within the AIDA project.
Verified and validated the functionalities of the AIDApy.
Supported the overall AIDA machine learning engine.

Communication of the AIDA activities and results to target audiences following the communications plan.
Ensured the exploitation of the AIDA outcomes following the exploitation plan.
Dissemination of the results of the project among the EC work programs and stakeholders of the project to create awareness and incited standarisation of AI efforts.
Promoted the scientific results of the AIDA project in the social media, traditional media and specialized newsletters.