Changing Learning rate

1

Hi, I am new to dl4j, i use version 1.0.0 beta3. i want to change the learning rate but i get the error “Cannot resolve method ‘learningRate’ in ‘Builder’”. How i can change it?

2

I barely have anything to go on here. It’s hard to tell if you have a version issue, what you’re trying to modify or how you’re editing your files.

Could you just start from our examples? Also, please use the latest version (beta7 now). I’m not sure how you chose beta3, but that’s pretty old at this point.
Also make sure whatever you’re trying to include in your project, all versions are the same.

3

thanks for replying. i tried with beta7 and your examples but i got the same error.

4

Here is my configuration code:
MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
.iterations(1)
.learningRate(0.1)
.rmsDecay(0.95)
.seed(12345)
.regularization(true)
.l2(0.01)
.weightInit(WeightInit.XAVIER)
.updater(new Nesterovs(0.1))
.list()
.layer(0, new GravesLSTM.Builder().nIn(iter.inputColumns())
.nOut(lstmLayerSize)
.activation(Activation.RELU6).build())

             .layer(1,	new	RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
                    .activation(Activation.SOFTMAX)		
                    .nIn(lstmLayerSize).nOut(nOut).build())
            .backpropType(BackpropType.TruncatedBPTT).tBPTTForwardLength(tbpttLength)
            .tBPTTBackwardLength(tbpttLength)
            .pretrain(true).backprop(true)
            .build();
5

i get the same error for “iterations()”, “rmsDecay()” and “regularization()”

How this hyper parameters can be changed???

6

Hi, you should probably first understand the concept of an “updater”:
https://deeplearning4j.konduit.ai/models/updaters

I would suggest just configuring the updater (which includes the learning rate and other parameters) directly.
Here’s a simple example of that:

I’m not sure where you’re getting the usage of iterations etc. That api is super old and was killed along time ago. Please stick to the modern examples. Half of your errors here are coming from using old examples and just guessing what might work rather than referencing something.
Try to slow down a bit.
I’m no

7

thanks a lot it helped :+1: