Sum Product Networks?

[hrstoyanov]

Ivan,
Congratulations, It was about time for someone to start a project like that!!! I have been investigating Aparapi for implementing Sum Product Networks, as SPNs seems to exhibit much better properties that RBMs or DBNs (you would not know it if you only listened to Prof. Hinton's classes:)

Questions:

1. Do you plan on implementing SPNs in addition to RBMs and DBNs?
2. Do you plan on switching to Java 8 soon? It seem that Java 8 lambdas+Aparapi are very suitable for the latest AMD Kaveri (and upcoming Berlin) APUs ?

Pozdravi,
Hristo Stoyanov

[ivan-vasilev]

Thank you very much for your interest! To answer your questions:

1. I was originally planning to implement Tensor Stacking Networks next. However SPNs look equally promising, so I might as well start with them.
2. Please check out the develop branch, which contains the latest sources (and it's relatively stable now), including Java 8 migration - the new lambda's and streams are used throughout the project. I have not yet switched to the latest version of Aparapi, which supports HSA and lambdas, but I plan to do so in the near future.

[hrstoyanov]

Thanks,
One last question: did you implement drop-out as part of the learning? This
seems to be a very effective trick.

/Hristo Stoyanov
On Apr 19, 2014 1:50 AM, "Ivan Vasilev" notifications@github.com wrote:

Thank you very much for your interest! To answer your questions:

1. I was originally planning to implement Tensor Stacking Networks next.
   However SPNs look equally promising, so I might as well start with them.
2. Please check out the _develop_https://github.com/ivan-vasilev/neuralnetworks/tree/developbranch, which contains the latest sources (and it's relatively stable now),
   including Java 8 migration - the new lambda's and streams are used
   throughout the project. I have not yet switched to the latest version of
   Aparapi, which supports HSA and lambdas, but I plan to do so in the near
   future.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/17#issuecomment-40864354
.

[ivan-vasilev]

I haven't implemented it yet, but I plan to do so.

[ivan-vasilev]

I just added experimental support for dropout in fully connected layers trained with backpropagation.

[hrstoyanov]

Thanks, I will check it out!