Title: Deep clustering: The use of the fuzzy k-means loss function in the simultaneous approach Authors:  Claudia Rampichini - University of Rome La Sapienza (Italy) [presenting]
Abstract: The proposed method combines the use of an autoencoder neural network with a fuzzy clustering algorithm. Deep clustering is a recent field of research that exploits the power of neural networks to improve clustering performance. The main idea is to use a deep neural network to reduce the input data's complexity and apply a clustering algorithm to the new reduced data. The proposed algorithm links the encoder part and the code space of a deep autoencoder neural network to a new layer. It jointly optimizes the model by minimizing the fuzzy k-means objective function. In addition, to avoid the problem of collapsing centres, a penalization term is added. The proposal shows margins of improvement compared with traditional clustering methods and the sequential deep clustering approach.