Manifold Learning for Real-World Event Understanding

Information coming from social media is vital to the understanding of dynamics involved in multiple events such as terrorist attacks and natural disasters. With spread popularization cameras means share content through networks, an event can be followed many different lenses vantage points. However, data present numerous challenges, frequently it necessary a great deal cleaning filtering techniques separate what related depicted contents otherwise useless. In previous effort ours, we decomposed into representative components aiming at describing details characterize its defining moments. lack minimal supervision guide combination somehow limited performance method. this paper, extend upon our prior work learning-from-data method for dynamically learning contribution more effective representation. The relies just few training samples (few-shot learning), which easily provided by investigator. obtained results on real-world datasets show effectiveness proposed ideas.