Julien Bastian

Julien Bastian

Applied mathematics master student

Works

Random Fourier Features for PAC-Bayesian Domain Adaptation

Building on PAC-Bayesian theory for Domain Adaptation, the research emphasizes kernel approximation methods with Random Fourier Features, to enhance data representation for classification in target domains without labels. We developped a novel PAC-Bayesian for this set-up and derived a learning algorithm.

Learning representation with Wavelets an application to ECG classification

In our research, we combined Convolutional Neural Networks and Continuous Wavelet Transformation techniques to enhance heartbeat classification using electrocardiograms. This combination aims to efficiently detect cardiac anomalies, like arrhythmia, offering potential advancements in cardiac health monitoring and diagnostics.