M
Michael P. Perrone
Researcher at IBM
Publications - 82
Citations - 2583
Michael P. Perrone is an academic researcher from IBM. The author has contributed to research in topics: Handwriting recognition & Handwriting. The author has an hindex of 21, co-authored 82 publications receiving 2492 citations. Previous affiliations of Michael P. Perrone include Brown University & GlobalFoundries.
Papers
More filters
When Networks Disagree: Ensemble Methods for Hybrid Neural Networks
TL;DR: Experimental results show that the ensemble method dramatically improves neural network performance on difficult real-world optical character recognition tasks.
Journal ArticleDOI
Cell Multiprocessor Communication Network: Built for Speed
TL;DR: The authors analyze the cell processor's communication network, using a series of benchmarks involving various DMA traffic patterns and synchronization protocols to illuminate this important point in multicore processor design.
Improving regression estimation: Averaging methods for variance reduction with extensions to general convex measure optimization
TL;DR: Experimental results are presented which demonstrate that the ensemble method dramatically improves regression performance on real-world classification tasks.
Proceedings ArticleDOI
Multicore Surprises: Lessons Learned from Optimizing Sweep3D on the Cell Broadband Engine
Fabrizio Petrini,G. Fossum,Juan C. Fernandez,Ana Lucia Varbanescu,N. Kistler,Michael P. Perrone +5 more
TL;DR: In the exploration to achieve the optimum level of performance for Sweep3D, the BE processor has enjoyed many pleasant surprises, such as a very high floating point performance, reaching 64% of the theoretical peak in double precision, and an over all performance speedup ranging from 4.5 times when compared with "heavy iron" processors, up to over 20 times with conventional processors.
Proceedings ArticleDOI
Pre-processing methods for handwritten Arabic documents
TL;DR: An orientation independent technique for baseline detection of Arabic words is described and it is shown how the baseline can be exploited for slope and skew correction before proceeding with the steps of line and word separation.