Understanding Minimum Probability Flow for RBMs Under Various Kinds of Dynamics
This work addresses training challenges for RBMs, a key model in unsupervised learning, but is incremental as it extends MPF with more general dynamics.
The paper tackles the problem of training Restricted Boltzmann Machines (RBMs) by investigating Minimum Probability Flow (MPF) learning as an alternative to contrastive divergence (CD), which has theoretical limitations, and shows that MPF outperforms CD in various RBM configurations.
Energy-based models are popular in machine learning due to the elegance of their formulation and their relationship to statistical physics. Among these, the Restricted Boltzmann Machine (RBM), and its staple training algorithm contrastive divergence (CD), have been the prototype for some recent advancements in the unsupervised training of deep neural networks. However, CD has limited theoretical motivation, and can in some cases produce undesirable behavior. Here, we investigate the performance of Minimum Probability Flow (MPF) learning for training RBMs. Unlike CD, with its focus on approximating an intractable partition function via Gibbs sampling, MPF proposes a tractable, consistent, objective function defined in terms of a Taylor expansion of the KL divergence with respect to sampling dynamics. Here we propose a more general form for the sampling dynamics in MPF, and explore the consequences of different choices for these dynamics for training RBMs. Experimental results show MPF outperforming CD for various RBM configurations.