Post by Rafael Sojo Garcia

Following up on our mission to bring robust AI solutions to the industrial sector, I am thrilled to announce our latest publication in the prestigious journal Expert Systems with Applications (ESWA)! 🚀 In my previous post, we tackled the challenge of deploying models with limited data using Transfer Learning. But what happens after deployment? In real-world industrial environments, machinery wears down and operating conditions inevitably change. This phenomenon, known as concept drift, can quickly degrade a model's reliability if left unaddressed. 📉 To solve this, we introduce LLR2, a fully unsupervised concept drift detector based on Bayesian networks. The best part? Once a drift is detected by LLR2, it adapts to the new conditions by directly applying the transfer learning methodology we developed in our previous work! This ensures the model stays accurate and up-to-date, without forgetting previously learned, still-valid information. 🏭 All the details are available in the full article 📄: https://lnkd.in/eMNfDUmk This milestone is another proud step in my Industrial PhD program (Doctorados Industriales DIN2024-013310, from the Ministry of Science and Innovation of Spain) at Aingura IIoT, and the result of a lot of hard work and collaboration. Therefore, I would like to thank: 👨‍🏫 My PhD advisors, Pedro Larrañaga and Concha Bielza, for their invaluable guidance at Universidad Politécnica de Madrid. 🏭 Javier Díaz Rozo, PhD and the entire team at Aingura IIoT, for providing the industrial datasets and sharing their practical expertise, which were fundamental to enriching this work. 💻 The Mobile and embedded-based HPC group at the Barcelona Supercomputing Center, for our fruitful discussions and their contribution to the computational optimization of our experimental framework. #MachineLearning #ConceptDrift #BayesianNetworks #IndustrialAI #IIoT #ConditionMonitoring #TransferLearning #ESWA #PhD