Francisco Mena

Promotional graphic for the “Code for Earth” programme. The design features a dark blue and teal background with abstract geometric shapes. Large text reads “Call for participation,” with the word “participation” inside a green oval with small circular accents. A green label at the top left shows “Phase 1” and the dates “24.02.2026 – 09.04.2026.” The top right displays the “Code for Earth” logo. Along the bottom are logos for ECMWF, the European Union, Copernicus, Destination Earth, and the European Weather Cloud.

📣 Applications are open for ECMWF’s Code for Earth 2026!

New data driven challenges across visualisation, machine learning, software development plus a brand new Africa focused stream with African partners.

📅 Apply by 9 April 2026

@codeforearth.bsky.social

www.ecmwf.int/en/about/med...

🎓 10 #PhD positions in #AI & #DataScience - #Berlin

BIFOLD is hiring 10 PhD candidates in:
🤖 #MachineLearning
🗄️ #DataManagement
🔗 #ML × #DM

Apply until Feb 13, 2026
www.jobs.tu-berlin.de/en/job-posti...

@tuberlin.bsky.social @rieck.mlsec.org
#AcademicSky #PhDSky #sciencejobs
#academicjobs

I'm traveling 🚄 towards Copenhagen 🇩🇰 for #Eurips. Happy to catch up if you are around 😀

We combine contrastive learning + modality-discriminative losses to structure features into shared and specific subspaces. Tested on four EO benchmarks (classification & regression) → consistent gains over both EO and ML state-of-the-art.

Multi-modal co-learning for Earth observation: enhancing single-modality models via modality collaboration - Machine Learning Multi-modal co-learning is emerging as an effective paradigm in machine learning, enabling models to collaboratively learn from different modalities to enhance single-modality predictions. Earth Obser...

I'm happy to share that our new paper in Multi-modal co-learning for Earth observation got published in the ML journal🌍📡
Here, we show how models trained on multiple sensor modalities can boost single-modality inference
🔗 link.springer.com/article/10.1...

Are these the happiest PhD students in the world? Nature - Brazil, Australia and Italy have the highest satisfaction scores in Nature’s global 2025 PhD survey — but are these nations really the best places to do a doctorate?

We asked 3,785 PhD students across 107 countries about their experiences. Where do you think the happiest doctoral candidates were?

go.nature.com/43usVmf

🌍 Excited to announce our Workshop on AI for Climate & Conservation (AICC) at #EurIPS2025 in Copenhagen! 🎉

📢 Call for Participation: sites.google.com/g.harvard.ed...

Confirmed speakers from Mistral AI, DeepMind, ETH Zurich, LSCE & more.

Looking forward to meeting and discussing in Copenhagen!

Working on representation learning for Earth Observation?
Come join the discussion at the EurIPS workshop "REO: Advances in Representation Learning for Earth Observation"

Call for papers deadline: October 15, AoE
Workshop site: sites.google.com/view/reoeurips

@euripsconf.bsky.social @esa.int

Always happy to receive those accepted paper email 😃

In the search for optimal multi-view learning models for crop classification with global remote sensing data Studying and analyzing cropland is a difficult task due to its dynamic and heterogeneous growth behavior. Usually, diverse data sources can be collect…

It is possible to reduce the number of experiments when searching for the best combination of encoder architecture and fusion strategy for crop classification 🌱🚜?

Spoiler alert: In our recent (open access) paper 📖, we show that it can!

www.sciencedirect.com/science/arti...

We are delighted to announce the #EurIPS 2025 Workshops 🎉: eurips.cc/workshops/

We received 52 proposals, which were single-blind reviewed by more than 35 expert reviewers, leading to 18 accepted workshops (acceptance rate 34.6%).

We are looking for an NLP postdoc/engineer to work on adding language capabilities to our Earth observation sensor-agnostic models (Atomizer, to be presented at BMVC25).

Details here: jobs.inria.fr/public/class...

Atomizer: arxiv.org/pdf/2506.13542
GEO-ReSeT project: anr.fr/Projet-ANR-2...

Snapshot of paper link

Did you know that mutual distillation can be used to make deep learning models robust to missing sensor data?
We present this in our recent paper from a collaboration between @dfki.bsky.social and Inria (evergreen team). Available at @ieeeaccess.bsky.social 🔓

ieeexplore.ieee.org/document/10994…

Considering the current substantial use of computational resources in deep learning research and its consequential impact on the carbon footprint 👣, it is important to look for systematic ways that lead us to reduce computational efforts

Instead of trying all possible combinations, the search could be reduced to a 2-step sequential search: 1) search for the best encoder architecture with early/input fusion, and then 2) with the encoder selected in (1), search for the best fusion strategy

When considering all the diverse encoder architectures (like convolutional or attention-based) and fusion strategies (like input and feature) from the literature, the search space of all possible model combinations is considerably big and a resource-wasting process.

In the search for optimal multi-view learning models for crop classification with global remote sensing data Studying and analyzing cropland is a difficult task due to its dynamic and heterogeneous growth behavior. Usually, diverse data sources can be collect…

It is possible to reduce the number of experiments when searching for the best combination of encoder architecture and fusion strategy for crop classification 🌱🚜?

Spoiler alert: In our recent (open access) paper 📖, we show that it can!

www.sciencedirect.com/science/arti...

🏞️ Today is @unep.org #WorldLakeDay!

Global, long-term satellite records developed by the ESA Climate Change Initiative shed light on lakes contribution to the hydrological, energy and carbon cycles and their response to climate change.

Check out the data set visualisations here: t1p.de/mvl0a

📣 Please share: We invite submissions to the 29th International Conference on Artificial Intelligence and Statistics (#AISTATS 2026) and welcome paper submissions at the intersection of AI, machine learning, statistics, and related areas. [1/3]

Also, don't hesitate to visit our CCS in Probabilistic Machine Learning for Earth Observation (TU2.M1)!
⏲️ Tuesday, 5 August, 10:30 - 11:45

Also, don't hesitate to visit our CCS in Probabilistic Machine Learning for Earth Observation (TU2.M1)!
⏲️ Tuesday, 5 August, 10:30 - 11:45

⏲️ Thursday, 7 August, 15:45 - 17:00
📜 On What Depends the Robustness of Multi-source Models to Missing Data in Earth Observation? in the TH4.P11: Multi-source Semantic Segmentation (oral 🎤)
⭐I'll present our findings about three major factors that drive the robustness to missing data sources.

⏲️ Tuesday, 5 August, 09:15 - 10:30
📜 A Multi-modal Co-learning Model with Shared and Specific Features for Land-cover Classification in the TUP1.PB: Cross-Domain Learning and Semantic Segmentation in RS (poster🖼️)
⭐ Here we leverage co-learning and multiple losses to improve single-modality inference

This coming week will be a thrilling and enriching experience at IGARSS 2025. I'll be presenting two works in multi-modal/source learning focused on missing data sources.

Let's catch up if you are around!

#IGARSS #IEEE #GRSS #AI4EO #EO #AI

During the last couple of years, we have read a lot of papers on explainability and often felt that something was fundamentally missing🤔

This led us to write a position paper (accepted at #ICML2025) that attempts to identify the problem and to propose a solution.

arxiv.org/abs/2402.02870
👇🧵

Academic "Deadlines"

GitHub - fmenat/DSensDp: Public repository of our research work at IEEE Access Public repository of our research work at IEEE Access - fmenat/DSensDp

The code is available at github.com/fmenat/DSensDp

We show that our multi-sensor approach is more robust in average than recent methods from the EO literature in three classification tasks, namely cropland classification, crop-type classification, and tree-species classification.

@interdonatos.bsky.social

Concretely, we use a mix of sensor dropout as data augmentation and mutual distillation to enhance collaborative learning across sensors, namely DSensD+. We leverage multi-task learning to combine various objectives to achieve an optimal robustness

Snapshot of paper link

Did you know that mutual distillation can be used to make deep learning models robust to missing sensor data?
We present this in our recent paper from a collaboration between @dfki.bsky.social and Inria (evergreen team). Available at @ieeeaccess.bsky.social 🔓

ieeexplore.ieee.org/document/10994…