Re:A TensorFlow implementation of the Differentiable Neural ...
by jostmey on Wednesday April 19, 2017 @12:21PM
(#14148931)
I hate the format of the Nature paper. I have no idea how the algorithm is different than their neural turing machine
[ Parent
| on HN ]
Re:A TensorFlow implementation of the Differentiable Neural ...
by kekeblom on Thursday April 20, 2017 @01:54AM
(#14154284)
The DNC has a similar structure as the neural turing machine. It has a controller network and a memory matrix with read and write heads. However, the DNC differs in it’s memory accessing mechanisms.
Firstly, the DNC has a memory allocation mechanism ensuring that the network can write to memory that is not being used. Secondly, it also contains free gates that allow the network to explicitly forget data stored in it’s memory. Thirdly, the DNC employs a temporal link matrix which tracks the order in which writes are made. The temporal link matrix allows the location based addressing mechanism to recover from a jump caused by a content based memory access.
Re:A TensorFlow implementation of the Differentiable Neural ...
by albertzeyer on Wednesday April 19, 2017 @02:25PM
(#14150346)
One extension is that the model can dynamically grow/reduce the external augmented memory.
Re:A TensorFlow implementation of the Differentiable Neural ...
by pavledjo on Wednesday April 19, 2017 @02:02PM
(#14150104)
IIRC, the difference is an additional read head. Also they removed location-based addressing for the read/write heads in favor of a combination of content-based adressing and a dynamic scheme to figure out empty locations.
Re:A TensorFlow implementation of the Differentiable Neural ...
by zardo on Wednesday April 19, 2017 @03:02PM
(#14150710)
Could you expand on the removal of location based addressing? That's not how I understood it, but... I don't understand it very well.
Re:A TensorFlow implementation of the Differentiable Neural ...
by hidden-markov on Wednesday April 19, 2017 @01:00PM
(#14149333)
Neural Turing machine is a Turing machine. This one is a differentiable random access memory machine.