Using Neural Network Toolbox

To use Neural Network Toolbox,

Q1. Is 'Parallel Computing Toolbox' nescessary to use 'Neural Network Toolbox' ?

 - I heard Parallel Computing Toolbox is only recommed for using Neural Network Toolbox.
   Which NN toolbox functions can't use without Parallel Computing Toolbox?

Q2. IS real GPU necessary to use Parallel Computing Toolbox?

 - Only GPU download is OK for using parallel Computing Toolbox, without real GPU?
 - Here 'GPU' in MATLAB means just general graphic card(mine is NVIDIA Quadro K5000)?
   or additional precess unit excepth general graphic card?

- Detail -

I'm trying to use 'Alexnet'. (My MATLAB is R2016b)

To do that, I set 'Neural Network Toolbox'

and there are some errors so I also set 'Parallel Computing Toolbox'

Now, there is a new error message "parallel.gpu.GPUDeviceManager.instance is not defined"

How can I solve this problem?

[Joss Knight]

What do you mean 'set Parallel Computing Toolbox'? Are you compiling some code to a stand-alone executable?

PCT is required to train a network in 16b, but not to use one for classification or prediction.

Answer 1

[Kojiro Saito]

A1. In R2016b, Parallel Computing Toolbox is recommended as written in this link but not mandatory. But some algorithm require Parallel Computing Toolbox. This answer will help you.

A2. If when you calculate your algorithm on GPU, you need real GPU. You can also find the GPU requirement on this link, in Parallel Computing Toolbox -> "Requirements for GPU computing". CUDA enabled GPU and CUDA driver are required.