type:: [[meeting]] external-links:: tags:: [[PROJECTS/PRIN-EMELIOT]] people:: date:: [[26-01-2023]] - 09:15 isconsortium:: true - # Agenda - **09:00-10:30 S3: ML-based development** - - Studies on ML-based development (Unisannio) - - Federated Learning (Polimi) - Approaches for code completion - LUPE (Univaq, collaboraiton with Unisannio) - Open discussion: studies on ML-based development, call for joint initiatives - 10:30-11:00 Coffee break - **11:00-12:30 - S4: Analysis and testing** - - chatbot testing (Unimib) - - intrusion detection (Unisannio) - Open discussion - Security analysis and testing of ML-intensive IoT - **12:30-13:00  - Organizzazione iniziative future, wrap up** - 13:00-14:00 Pranzo - - ## Studies on ML-based development (Unisannio) - [Open Challenges in Federated Machine Learning | IEEE Journals & Magazine | IEEE Xplore (cineca.it)](https://ieeexplore-ieee-org.univaq.clas.cineca.it/document/9829276) - ## Claudio - [LUPE.pptx](../assets/LUPE_1674725870195_0.pptx) - ## Mastropaolo [[people/Mastropaolo]] - [[people/Mastropaolo]] - RSSEs - Rule-based (prima generazione) - Soffrono di problemi di generalizability - Es. JSummarizer usa template predefiniti progettati per Java - Data-driven techniques - Performance non eccezionali. Sono 3% di metodi viene summarized correttamente oppure solo il 9% di codice buggy viene fixato in modo corretto. - Prossimo step: Investigare pre-trained models - BERT (Google) - T5 (Google) - text-to-text transfer tranformer - Text as Input and Text as Output - DONE Vedere [Studying the Usage of Text-To-Text Transfer Transformer to Support Code-Related Tasks | IEEE Conference Publication | IEEE Xplore (cineca.it)](https://ieeexplore-ieee-org.univaq.clas.cineca.it/document/9401982) id:: 648b0617-8e0d-49a4-81dd-7bf3576cbd5b - DONE [Using deep learning to generate complete log statements | Proceedings of the 44th International Conference on Software Engineering (cineca.it)](https://dl-acm-org.univaq.clas.cineca.it/doi/10.1145/3510003.3511561) id:: 648b0617-6649-412a-a148-68f04c0be187 - [An Empirical Study on the Usage of Transformer Models for Code Completion | IEEE Journals & Magazine | IEEE Xplore (cineca.it)](https://ieeexplore-ieee-org.univaq.clas.cineca.it/document/9616462) id:: 648b0617-cf41-4390-b192-d90855c19e31 - RoBERTA (Facebook) - *Pretraining then fine-tuning* - ## Chatbot testing - [chatbot2.pptx](../assets/chatbot2_1674729106591_0.pptx) - - - ## Intrusion detection - - 15-16 Giugno - Prossimo meeting EMELIOT a Salerno (pomeriggio del 15 e mattina del 16) - Lun-Mar --> Sattose - Martedi, ??? --> Scuola estiva -