mdCNN: Multidimensional CNN library in Matlab

Hi, there is a 1D example under the Demo folder, you can try to follow it. On Wednesday, August 19,

Hi, the current optimizer is SGD There is no support for Adam, but I guess you can modify '

From the config file, you can control whats the last layer type. Please refer to the documentation On

Hi Hanh, thanks. The receptive field for a single neuron is a function of its layer parameters -

Hi, Sadani, Please refer to the doc file under "classify a single sample" there is a sample

Hi I am running cifar-10 demo. actually it takes very long time. for me, I run in 12 hours but just

Currently no, pooling layer can only come after convolutional layer. And average pooling is not

First, you should remove the softmax layer and change the output layer to use the MSE cost function.

Happy to hear it works well :) For regression, you can take the last layer outputs as your prediction

Thank you Hagay for all your help On Wednesday, April 10, 2019 at 2:13:42 AM UTC-5, Hagay Garty wrote

Hi, you can follow the doc file in the main git folder. Section 5 - 'Train a network' Another

learning rate starts from ni_initial and after 'noImprovementTh' consecutive iterations with

Hello again, thanks for your quick response, It looks like the version that i had was the one from

Thank you.It successfully worked.

Thank you. I will try the latest version.

I am not alucky person, i will try to go through document file and this group. Later, i may ask you

latest version solves the bug Thanks S On Tuesday, November 27, 2018 at 5:11:28 PM UTC+2, Hagay Garty

Hi, batch normalization layer is now supported in the latest code. On Monday, April 23, 2018 at 8:15:

In the new version, you can set the regression layer loss function to a user-defined function or use

Thanks 在 2018年10月5日星期五 UTC+1下午2:02:41，Hagay Garty写道： Hi, the kernels are in the same dimensions of

Hi, to classify a single image you can use the below code: image=dataset.I{1}; outs = feedForward(net

Yes you can use 2018a

Not sure what is your issue. You can try without normalization, did you try to set '

You can disable the backprop verification by adding net.runInfoParam.verifyBP = 0; to your config

Implementing CNN using FPGA is not a simple task Moreover, converting a matlab toolbox directly to

maxpooling works on the output of the network layer. The order is 1) convolution, 2) stride, 3) max

Hi, yes you can do both. Autoencoders can be done using a specific network structure (specified in

Thank you. I understood. Best regard Hang-Rai Kim 2018년 6월 2일 토요일 오전 1시 0분 10초 UTC+9, Hagay Garty 님의

You are right the scesRate is calculated on the test data only, after each training iteration.

Yes, simply call Train again using the 'net' variable returned from previous Train call. On