Novel tree-ensemble based methods for weakly-supervised structured output prediction with applications in biomedicine KU Leuven
Recent studies in structured output prediction, an umbrella term for machine learning tasks where multiple outputs must be predicted, have identified the veracity of the data as a major challenge. More specifically, structured output prediction datasets present noise in the output space due to faulty equipment, high cost of annotation or high volume of data, meaning that they are weakly-supervised. State-of-the-art methods, however, often ...