horn-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Edward J. Yoon" <edward.y...@samsung.com>
Subject Just share my current status
Date Wed, 27 Jul 2016 07:54:06 GMT
Hi forks,

I implemented and tested the dropout neural net. 20 tasks (synchronous
parallel SGD) used, and training time is about 30 mins, reached 97~8%
accuracy on my PC[1]. Recently I'm writing CNN: HORN-10. With this, I hope
we can achieve 99+% accuracy of MNIST example. :-)

At the same time, I'm thinking about parallel regularization method on
distributed computing environments: HORN-27. I hope this can be a
different/strength point from others layer abstraction model.

Thanks~

1. https://github.com/edwardyoon/incubator-horn/blob/master/README.md

--
Best Regards, Edward J. Yoon




Mime
View raw message