Previously, SustainML kept task results, Hugging Face searches, and model comparisons in memory for a single session. Closing or restarting the app meant losing that work. You can now save results, searches, or comparisons to a file you name and reopen it later, even after a restart. To keep your work, click Save and choose a name. Nothing is saved automatically.
Machine learning models can achieve impressive classification results while providing little insight into how they internally organize and distinguish information. Latent Boost addresses this challenge by incorporating the structure of a model's latent representations directly into the training objective. The approach encourages representations belonging to the same class to form compact clusters while separating different classes, improving interpretability and accelerating convergence with minimal additional cost.
We are proud to share that the paper “Latent Boost: Enhancing Interpretability Through Loss-Defined Classification Objective in Structured Latent Spaces” has received a Best Paper Award.
Monitoring the condition of bridges, roads, tunnels, and other critical infrastructure is essential for ensuring public safety and preventing costly failures. However, developing reliable AI solutions for this task has been limited by the scarcity of high-quality annotated data.
Artificial intelligence systems continue to evolve rapidly, but this progress is often accompanied by rising computational requirements and energy consumption. Training and deploying modern machine learning models can require substantial hardware resources, creating challenges related to scalability, operational cost, and environmental impact.
SustainML is among these nine innovative projects dedicated to creating a sustainable ML framework for Green AI.
Plaza de la Encina 10-11, Núcleo 4, 2ª Pl.
28760 Tres cantos - Madrid (España)
This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101070408.
