How to acquire new knowledge from new added training data while retaining the knowledge learned before is an important problem for incremental learning. In order to handle this problem, we propose a novel algorithm that enables support vector machines to accommodate new data, including samples that correspond to previously unseen classes, while it retains previously acquired knowledge. Furtherm...

To prevent process interruption and eventual losses, the need for a reliable fault detection and diagnosis system (FDD) is completely acknowledged. Besides the capability to recognize known faults automatically, a further requirement for a FDD is adaptability. If the model cannot be adapted to deal with changes, variations due to external changes, decaying performance, Poisoning of catalyst etc...

This work presents the application of a new, enhanced version of the incremental learning system INTHELEX (INcremental THEory Learner from EXamples), the learning component in the architecture of the EU project COLLATE, dealing with the annotation of cultural heritage documents. Due to the complex shape of the handled material, the addition of multistrategy capabilities was needed to improve th...

3D object classification has attracted appealing attentions in academic researches and industrial applications. However, most existing methods need to access the training data of past classes when facing common real-world scenario: new objects arrive a sequence. Moreover, performance advanced approaches degrades dramatically for learned (i.e., catastrophic forgetting), due irregular redundant g...

Deep learning architectures have shown remarkable results in scene understanding problems, however they exhibit a critical drop of performances when are required to learn incrementally new tasks without forgetting old ones. This catastrophic phenomenon impacts on the deployment artificial intelligence real world scenarios where systems need and different representations over time. Current appro...

Incremental learning from noisy data presents dual challenges: that of evaluating multiple hypotheses incrementally and that of distinguishing errors due to noise from errors due to faulty hypotheses. This problem is critical in such areas of machine learning as concept learning, inductive programming, and sequence prediction. I develop a general, quantitative method for weighing the merits of ...