Deep Clustering with a Dynamic Autoencoder: From Reconstruction towards Centroids Construction

23 Jan 2019  ยท  Nairouz Mrabah, Naimul Mefraz Khan, Riadh Ksantini, Zied Lachiri ยท

In unsupervised learning, there is no apparent straightforward cost function that can capture the significant factors of variations and similarities. Since natural systems have smooth dynamics, an opportunity is lost if an unsupervised objective function remains static during the training process. The absence of concrete supervision suggests that smooth dynamics should be integrated. Compared to classical static cost functions, dynamic objective functions allow to better make use of the gradual and uncertain knowledge acquired through pseudo-supervision. In this paper, we propose Dynamic Autoencoder (DynAE), a novel model for deep clustering that overcomes a clustering-reconstruction trade-off, by gradually and smoothly eliminating the reconstruction objective function in favor of a construction one. Experimental evaluations on benchmark datasets show that our approach achieves state-of-the-art results compared to the most relevant deep clustering methods.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Clustering Fashion-MNIST DynAE Accuracy 0.591 # 12
NMI 0.642 # 10
Image Clustering MNIST-full DynAE NMI 0.964 # 3
Accuracy 0.987 # 3
Image Clustering MNIST-test DynAE NMI 0.963 # 1
Accuracy 0.987 # 1
Image Clustering USPS DynAE NMI 0.948 # 2
Accuracy 0.981 # 2

Methods