Semi-unsupervised Learning of Human Activity using Deep Generative Models
This addresses a practical issue for real-world datasets like human activity data, but it is incremental as it builds on prior deep generative models.
The paper tackles the problem of semi-unsupervised learning, where some classes are sparsely labeled and others unlabeled, by introducing a Gaussian mixture deep generative model that achieves superior classification performance on MNIST compared to an existing model.
We introduce 'semi-unsupervised learning', a problem regime related to transfer learning and zero-shot learning where, in the training data, some classes are sparsely labelled and others entirely unlabelled. Models able to learn from training data of this type are potentially of great use as many real-world datasets are like this. Here we demonstrate a new deep generative model for classification in this regime. Our model, a Gaussian mixture deep generative model, demonstrates superior semi-unsupervised classification performance on MNIST to model M2 from Kingma and Welling (2014). We apply the model to human accelerometer data, performing activity classification and structure discovery on windows of time series data.