We consider the problem of spectral clustering under group fairness
constraints, where samples from each sensitive group are approximately
proportionally represented in each cluster. Traditional fair spectral
clustering (FSC) methods consist of two consecutive stages, i.e., performing