BEGIN:VCALENDAR VERSION:2.0 PRODID:-//datacraft - ECPv6.16.3//NONSGML v1.0//EN CALSCALE:GREGORIAN METHOD:PUBLISH X-WR-CALNAME:datacraft X-ORIGINAL-URL:https://datacraft.paris X-WR-CALDESC:Events for datacraft REFRESH-INTERVAL;VALUE=DURATION:PT1H X-Robots-Tag:noindex X-PUBLISHED-TTL:PT1H BEGIN:VTIMEZONE TZID:Europe/Paris BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:20210328T010000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:20211031T010000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:20220327T010000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:20221030T010000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:20230326T010000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:20231029T010000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:20240331T010000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:20241027T010000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:20250330T010000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:20251026T010000 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTART;TZID=Europe/Paris:20240228T180000 DTEND;TZID=Europe/Paris:20240228T190000 DTSTAMP:20260601T183018 CREATED:20240131T142206Z LAST-MODIFIED:20241205T132815Z UID:9887-1709143200-1709146800@datacraft.paris SUMMARY:State of the art - Practical insights for LLM fine-tuning and evaluation DESCRIPTION:Inscription\n \n \n \n \n \n \n \n \n [The workshop will be dispense in english ]  \nMachine Learning Level : \n**Good knowledge of Machine Learning \n Python Level \n*Basic skills in Python \nWorkshop description \nIf vendors almost announce to sell AGI-as-as-service through API to enterprise client\, the reality after trying to use proprietary LLM as a service on a specific use case is often different. Non relevant generation\, out-of-context answer\, misunderstanding of the user queries and more broadly lacking subject matter expertise starts to erode the users and shareholders confidence of the potential transformative power of deploying LLM-enhanced business workflow across your organisation. You’re not alone in this journey. \nIn this talk\, we’ll explore the landscape of fine-tuning solutions for open-source LLM\, weighing their pros and cons. We’ll delve into the data required and how to design a robust evaluation framework to systematically assess your in-house model’s performance. \nWe’ll deep dive on the subtle differences between the Parameter Efficient Finetuning Methods PEFT)\, the reinforcement learning approaches\, what to keep in mind when considering which one to use. \nThis talk is a synthesis of deploying LLM capabilities at various organisations\, from startup to corporate environments. It’s a blend of insights from research papers and pragmatic experiences. We won’t go onto the details of the mathematical operations under the hood for each fine-tuning approach\, instead our goal is to share the intuition of those concepts\, equipping you to design an effective roadmap for fine-tuning an LLM for your specific business use case. \nSlides in pdf will be made available for free on the speaker twitter at the end of the talk @fpaupier. \nIntervenants : \nFrançois Paupier\, machine learning engineer\, fpaupier engineering services \n  URL:https://datacraft.paris/event/etat-de-lart-from-agi-promises-to-llm-realities-practical-insights-into-language-model-fine-tuning-and-evaluation/ CATEGORIES:- Event in English - END:VEVENT BEGIN:VEVENT DTSTART;TZID=Europe/Paris:20231011T100000 DTEND;TZID=Europe/Paris:20231011T180000 DTSTAMP:20260601T183018 CREATED:20230609T074729Z LAST-MODIFIED:20231006T121251Z UID:8725-1697018400-1697047200@datacraft.paris SUMMARY:PARIS WORKSHOP - Frugal AI techniques applied to Image Semantic Segmentation DESCRIPTION:Inscription\n \n \n \n \n \n \n \n \n Machine Learning prerequisites** : Good skills \nPython prerequisites** : Good skills \nTechnical prerequisitesBring your own laptop \nSpeakersThis workshop will be animated by : \n\nDenis Marraud\, Image Processing Senior Expert\, Airbus Defence and Space\n\nWorkshop overviewFollowing the Frugal AI overview workshop organized in May\, this next workshop will aim at putting in practice and comparing the interest of some relevant Frugal AI techniques to solve a domain adaptation problem for one or several image semantic segmentation tasks. The cost and complexity of dense annotations required for this task makes it particularly interesting to leverage any technique contributing to the reduction of the number of required manual annotations.Envisaged techniques may include advanced data augmentation and pseudo-labelling methods\, weakly supervised or self-supervised methods. Various domain adaptation tasks may be considered : either domain specialization (or transductive learning) where the test domain is known in advance and closed\, domain extension where the performance should be optimized for both source and target domain\, or domain adaptation where the performance should be optimized on the target domain only. \nDataset descriptionExtracts from publicly available image segmentation datasets will be delivered to the participants. These datasets will cover various application domains such as medical imagery\, satellite or aerial imagery and self driving car imagery. \nAlgorithmic methodsAmong the techniques that could be tested during this workshop : \n\nAdvanced data augmentation methods (based on existing libraries)\nAdvanced pseudo-labelling methods (making use of non annotated data)\nSelf-supervised methods (e.g. contrastive learning\, predictive learning) as a pre-training step\nTest time adaptation methods to adapt to local context URL:https://datacraft.paris/event/frugal-ai-techniques-applied-to-image-semantic-segmentation/ LOCATION:datacraft –\, 3 rue Rossini\, 75009 Paris\, France CATEGORIES:- Event in English -,on-site event ORGANIZER;CN="datacraft":MAILTO:contact@datacraft.paris END:VEVENT BEGIN:VEVENT DTSTART;TZID=Europe/Paris:20230926T170000 DTEND;TZID=Europe/Paris:20230926T190000 DTSTAMP:20260601T183018 CREATED:20230707T081201Z LAST-MODIFIED:20230719T115740Z UID:8885-1695747600-1695754800@datacraft.paris SUMMARY:WORKSHOP - Continuous estimation of the environmental impact of Machine Learning solutions : methodology\, best practices and tools DESCRIPTION:Inscription\n \n \n \n \n \n \n \n \n Machine Learning prerequisites* : Basic knowledge in ML/Data/IA \nPython prerequisites* : Basic knowledge in Python \nTechnical prerequisitesNone \nSpeakersThis workshop will be presented by Samuel Rincé\, Lead Data Scientist at Alygne & OS Contributor at Boavizta \nWorkshop overview \nAlthough digital technologies currently represent about 2.1-3.9% of greenhouse gas emissions worldwide\, they are likely to double by 2030. Thus\, in a world where carbon emissions\, energy and water usage are becoming major concerns\, rethinking how we develop IT projects\, such as AI and ML projects\, to aim at reducing our digital environmental footprint is a real challenge. \nHowever\, there is no framework to take these issues into account.How can we integrate environmental footprint assessment into AI projects?From development to production\, how can we manage environmental impacts?What tools are needed to simplify and automate measurement? \nAfter this workshop\, you’ll be able to think about more sustainable approaches to your AI projects. \nCourse of the workshop \n\nMethodology for environmental impact estimation\nImpact evaluation tools in the development phase\nValidation and ecological benchmarking in CI\nImpact monitoring in production\nPractical project use cases URL:https://datacraft.paris/event/continuous-estimation-of-the-environmental-impact-of-ml-solutions/ LOCATION:datacraft –\, 3 rue Rossini\, 75009 Paris\, France CATEGORIES:- Event in English - ORGANIZER;CN="datacraft":MAILTO:contact@datacraft.paris END:VEVENT BEGIN:VEVENT DTSTART;TZID=Europe/Paris:20230831T190000 DTEND;TZID=Europe/Paris:20230831T200000 DTSTAMP:20260601T183018 CREATED:20230726T154032Z LAST-MODIFIED:20230728T123754Z UID:8976-1693508400-1693512000@datacraft.paris SUMMARY:STATE-OF-THE-ART - Frugal AI: Knowledge Extraction for Species Description DESCRIPTION:Inscription\n \n \n \n \n \n \n \n \n Machine Learning prerequisites* : Basic knowledges in ML/IA/data \nPython prerequisitesNone \nTechnical prerequisitesNone \nSpeakersThis workshop will be animated by : \n\nMaya Sahraoui\, PhD student at Sorbonne University\, ISIR and MNHN\n\nWorkshop overviewJoin us for a presentation of the latest research work of Maya\, on the potential of frugal AI. \nWith the constraint of few annotated data\, Maya has adapted a promising self-training method\, that is the teacher-student architecture\, to confidently propagate pertinent annotations to new unlabelled data. And she devised a comprehensive test protocol to assess the annotation. What she will present can be applied in many fields (marketing\, maintenance…) where you have to create a knowledge base from a few annotated datasets. \nFor more details:Maya focuses on elevating knowledge extraction models for analyzing biological species descriptions. Her research introduces a distantly supervised model for Named Entity Recognition (NER) and outlines a robust protocol for constructing knowledge graphs from entity labeling.To ensure rigorous evaluation\, she has meticulously devised a comprehensive test protocol consisting of two datasets. The first dataset includes entities encountered during training\, while the second comprises entirely new entities\, challenging the limits of her models.Throughout her investigation\, she encountered two significant scientific challenges: specificity of vocabulary and turn-of-phrase\, as well as missing annotations. She is excited to share how she tackled these hurdles by proposing a language model pre-training technique to enhance NER precision on both datasets. Moreover\, she will demonstrate the efficacy of our teacher-student architecture\, formulated as self-training\, which achieved remarkable recall on both test sets.Her findings shed light on the indispensable role of recent language models in deciphering complex and specialized texts. The implications are far-reaching\, benefiting researchers in species diversity and evolution\, and offering potential applications in comparative morphology and biodiversity informatics. \n  \n\n\n\n\n\n\n\n\n\n\n\n\n\n\n \n\n\n\n\n\n\n\n\n\n\n\n\n\n\n  URL:https://datacraft.paris/event/state-of-the-art-frugal-ai-knowledge-extraction-for-species-description/ LOCATION:datacraft –\, 3 rue Rossini\, 75009 Paris\, France CATEGORIES:- Event in English - ORGANIZER;CN="datacraft":MAILTO:contact@datacraft.paris END:VEVENT BEGIN:VEVENT DTSTART;TZID=Europe/Paris:20220624T110000 DTEND;TZID=Europe/Paris:20220624T120000 DTSTAMP:20260601T183018 CREATED:20220524T161948Z LAST-MODIFIED:20220524T162813Z UID:6825-1656068400-1656072000@datacraft.paris SUMMARY:MINDSHAKE TIME - Vision Transformer classification applied to computer vision medical diagnosis DESCRIPTION:inscription\n \n \n \n \n \n In this workshop\, we will discuss on Vision Transformer classification applied on medical diagnosis. After introducing Transformers and the mechanism of self-attention\, we will show the results of the ViT architecture compared with a classic CNN and a multistage architecture composed of successive CNNs. \nWe will then show how In recent years\, the scientific community focused on developing Computer-Aided Diagnosis tools that could improve clinicians’ bone fracture diagnosis\, primarily based on Convolutional Neural Networks (CNNs). However\, the discerning accuracy of fractures’ subtypes was far from optimal. The aim of this study is to evaluate a new CAD system based on Vision Transformers (ViT) and to assess whether clinicians’ diagnostic accuracy could be improved using this system. \nTo demonstrate this\, we will discuss an evaluation made by 11 clinicians\, who were asked to classify 150 proximal femur fracture images with and without the help of the ViT.  \nWorkshop led by Leonardo Tanzi\, PhD student at Polytechnic University of Turin URL:https://datacraft.paris/event/mindshake-time-vision-transformer-classification-applied-to-computer-vision-medical-diagnosis/ LOCATION:datacraft –\, 3 rue Rossini\, 75009 Paris\, France CATEGORIES:- Event in English - ORGANIZER;CN="datacraft":MAILTO:contact@datacraft.paris END:VEVENT BEGIN:VEVENT DTSTART;TZID=Europe/Paris:20220510T180000 DTEND;TZID=Europe/Paris:20220510T190000 DTSTAMP:20260601T183018 CREATED:20220411T082457Z LAST-MODIFIED:20220415T133510Z UID:6648-1652205600-1652209200@datacraft.paris SUMMARY:Mindshake Time - A Framework to Learn with Interpretation DESCRIPTION:inscription\n \n \n \n \n \n Workshop led by Jayneel Parekh\, Pavlo Mozharovskyi\, Florence d’Alché-Buc\, Télécom Paris \nWhat’s a Mindshake Time ?Mindshake Time is designed as a discussion circle on a given research topic. It allows to discuss the latest advances\, share thoughts\, and initiate collaborations. It is intended for all experts who wish to broaden or confront their knowledge and vision\, at the cutting edge of the state of the art. \nTheme of this event :To tackle interpretability in deep learning\, we present a novel framework that jointly learns a predictive model and its associated interpretation model. The interpreter provides both local and global interpretability about the predictive model in terms of human-understandable high level attribute functions\, with minimal loss of accuracy. This is achieved by a dedicated architecture and well chosen regularization penalties. A detailed pipeline to visualize the learnt features is also developed. Moreover\, besides generating interpretable models by design\, our approach can be specialized to provide post-hoc interpretations for a pre-trained neural network. We validate our approach against several state-of-the-art methods on multiple datasets and show its efficacy on both kinds of tasks. \nPaper’s link : https://arxiv.org/abs/2010.09345 \nPrerequisites : Basic knowledge in deep learning URL:https://datacraft.paris/event/mindshake-time-a-framework-to-learn-with-interpretation/ LOCATION:datacraft –\, 3 rue Rossini\, 75009 Paris\, France CATEGORIES:- Event in English - ORGANIZER;CN="datacraft":MAILTO:contact@datacraft.paris END:VEVENT END:VCALENDAR