LIRS: Enabling efficient machine learning on NVM-based storage via a lightweight implementation of random shuffling
This work addresses a performance bottleneck in machine learning training systems, offering incremental improvements for researchers and practitioners using NVM-based storage.
The paper tackled the problem of inefficient random shuffling of training data in machine learning due to slow random accesses in hard disk drives, proposing LIRS, a lightweight implementation for non-volatile memory-based storage, which reduced training time by up to 49.9% for SVM and 43.5% for DNN while improving DNN testing accuracy by 1.01%.
Machine learning algorithms, such as Support Vector Machine (SVM) and Deep Neural Network (DNN), have gained a lot of interests recently. When training a machine learning algorithm, randomly shuffle all the training data can improve the testing accuracy and boost the convergence rate. Nevertheless, realizing training data random shuffling in a real system is not a straightforward process due to the slow random accesses in hard disk drive (HDD). To avoid frequent random disk access, the effect of random shuffling is often limited in existing approaches. With the emerging non-volatile memory-based storage device, such as Intel Optane SSD, which provides fast random accesses, we propose a lightweight implementation of random shuffling (LIRS) to randomly shuffle the indexes of the entire training dataset, and the selected training instances are directly accessed from the storage and packed into batches. Experimental results show that LIRS can reduce the total training time of SVM and DNN by 49.9% and 43.5% on average, and improve the final testing accuracy on DNN by 1.01%.