Parallelized Training of Restricted Boltzmann Machines using Markov-Chain Monte Carlo Methods
This work addresses the training time bottleneck for RBMs in recommendation systems, making them more practical, though it is incremental as it applies existing parallelization methods to a known problem.
The paper tackled the computationally expensive training of Restricted Boltzmann Machines (RBMs) for recommendation systems by applying distributed parallel training using the Horovod framework, reducing training time from 5 hours on a single CPU node to just over 12 minutes on 64 CPU nodes.
Restricted Boltzmann Machine (RBM) is a generative stochastic neural network that can be applied to collaborative filtering technique used by recommendation systems. Prediction accuracy of the RBM model is usually better than that of other models for recommendation systems. However, training the RBM model involves Markov-Chain Monte Carlo (MCMC) method, which is computationally expensive. In this paper, we have successfully applied distributed parallel training using Horovod framework to improve the training time of the RBM model. Our tests show that the distributed training approach of the RBM model has a good scaling efficiency. We also show that this approach effectively reduces the training time to little over 12 minutes on 64 CPU nodes compared to 5 hours on a single CPU node. This will make RBM models more practically applicable in recommendation systems.