Efficient Probabilistic Latent Semantic Indexing using Graphics Processing Unit

Eli Koffi Kouassi, Toshiyuki Amagasa, Hiroyuki Kitagawa
2011 Procedia Computer Science  
In this paper, we attempt to accelerate the Probabilistic Latent Indexing (PLSI) exploiting the high parallelism of Graphic Processing Unit (GPU). Our proposal is composed of three methods. The first method is to accelerate the Expectation-Maximization (EM) computation by applying GPGPU matrix-vector multiplication. The second method uses the same principles as the first method but deals with the sparseness of co-occurrence of words and documents. The third method is to use the concurrent
more » ... execution, which is available on NVIDIA Fermi architecture, in order to speed up the second method. We compare the results to the most recent parallel execution of PLSI which combines a method of parallelization by OpenMP with the Message Passing Interface (MPI) for distributed memory parallelization. The experiments show that our method could be more than 100 times faster than the previous results. By dealing with the sparseness of the data, we could not only process more documents and words using GPU, but we could also keep more data on the device memory so that we can avoid massive data copy transfer between the host and the device susceptible to reduce the execution performance.
doi:10.1016/j.procs.2011.04.040 fatcat:jxavxqszonawtpqdsdp6xebfje