BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.4.0.1//EN TZID:Europe/Paris X-WR-TIMEZONE:Europe/Paris BEGIN:VEVENT UID:343@isdm.umontpellier.fr DTSTART;TZID=Europe/Paris:20231130T140000 DTEND;TZID=Europe/Paris:20231130T140000 DTSTAMP:20260828T084751Z URL:https://isdm.umontpellier.fr/events/phytosociology-meets-artificial-in telligence-accurate-habitat-type-prediction-based-on-deep-learning/ SUMMARY:Phytosociology meets artificial intelligence: accurate habitat type prediction based on deep learning DESCRIPTION:Room 109\, Building 9\, St Eloi campus\n\nMachine Learning in M ontpellier\, Theory &\; Practice\n\nBiodiversity is under severe pressu re\, as many different disturbance events threaten terrestrial and marine ecosystems with varying impacts. Therefore\, habitat distribution modellin g\, which aims to quantify the statistical links between environmental cov ariates and an habitat’s occurrence\, is increasingly relevant. Herein\, we present two different approaches to guide investment\, management and regulatory decisions. Firstly\, a framework based on tabular data\, which experiments with different network architectures\, feature encodings\, hyp erparameter tuning and noise addition strategies to identify the optimal m odel for habitat classification based on plant species composition. Second ly\, we introduce Pl@ntBERT\, which leverages sophisticated natural langua ge processes based on transformers (i.e.\, models with attention component s able to learn contextual relations between categorical and numerical fea tures). In particular\, since they reinforce each other\, the pipeline mak es use of both masked language modelling and text classification. The firs t step helps to get a statistical understanding of the plant species compo sition (the language in which the model is trained in). Then\, subsequent training is used to assign an habitat type to sentences describing vegetat ion plots. The fine-tuning of a pretrained foundation model on in-domain d ata shows significant upgrade. Notably\, it clearly outperforms previous s tate-of-the-art methods by pushing the accuracy score on a large database containing millions of European samples. Finally\, our results showcase th at flora is a strong marker of habitat type and doesn'\;t need to be c oupled with environmental spatial data to train neural networks with high predictive power. Looking forward to seeing you.\n\nMachine Learning in Mo ntpellier\, Theory &\; Practice ATTACH;FMTTYPE=image/jpeg:https://isdm.umontpellier.fr/wp-content/uploads/ 2026/06/ml-mtp-gC78d5.png CATEGORIES:ML MTP LOCATION:Saint Eloi Campus\, Building 9\, Room 109\, Montpellier\, X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Montpellier\, ;X-APPLE-RADI US=100;X-TITLE=Saint Eloi Campus\, Building 9\, Room 109:geo:0,0 END:VEVENT BEGIN:VTIMEZONE TZID:Europe/Paris X-LIC-LOCATION:Europe/Paris BEGIN:STANDARD DTSTART:20231029T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET END:STANDARD END:VTIMEZONE END:VCALENDAR