Error : The prediction sequences are of feature dimension 1 but the input layer expects sequences of feature dimension 4

Question

nuri ahn 2019-8-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/477281-error-the-prediction-sequences-are-of-feature-dimension-1-but-the-input-layer-expects-sequences-of

回答： Benjamin Ballintyn 2019-9-5

net_0.mat

在 MATLAB Online 中打开

I made this simple network for multiple inputs and single output regression.

This is the code I used. I attached the model file.(net_0.mat)

%Question
Question =  [0.3 0.4623115577889447 0.591304347826087 4.605170185988092];
obsDia_norm = Question(1);
obsCenterX_log = Question(2);
obsCenterY_log = Question(3);
freq_log = Question(4);
load net_0 %please download it from the attachment.
analyzeNetwork(net_0) 
assembledNet = assembleNetwork(net_0)
XTest = Question;
YPred = predict(assembledNet,XTest);

It shows this error.

Error using nnet.internal.cnn.util.NetworkDataValidator/assertValidSequenceInput (line 427)
The prediction sequences are of feature dimension 1 but the input layer expects sequences of
feature dimension 4.
Error in nnet.internal.cnn.util.NetworkDataValidator/validateDataForPredict (line 83)
                this.assertValidSequenceInput( X, inputSize );
Error in DAGNetwork/predictRNN (line 7)
dataValidator.validateDataForPredict(data, dispatcher, this.NetworkInfo);
Error in DAGNetwork/predict>predictBatch (line 245)
    Y = net.predictRNN(X, dispatcher, precision, predictNetwork, gpuShouldBeUsed);
Error in DAGNetwork/predict (line 118)
    Y = predictBatch( ...
Error in Main2_Restore_and_Estimation (line 23)
YPred = predict(assembledNet,XTest);

Q1. I don't understand this error message :

The prediction sequences are of feature dimension 1 but the input layer expects sequences of
feature dimension 4.

What should I change?

Q2. I'm not sure if it's right to use sequence layer.

In my data set, the four inputs are not a sequence. they are just independant variables.

(exactly the four inputs are diameter, x position, y position and frequency.)

But I can't find the other option. As I found there are just two input layers : sequenceInputLayer and imageInputLayer.

Is it okay to use sequenceInputLayer in this case?

Q3. I just want to get an output as a result of the 'FC_out' layer. I don't want to get a MSE from the regression output layer.

But If I remove the regression output layer, it shows me this error.

What can I use to get a output of 'FC_out' layer?

Caused by:
    Network: Missing output layer. The network must have one output layer.
    Layer 'FC_out': Unused output. Each layer output must be connected to the input of another
    layer.
    Layer 'Input': Empty AverageImage property. For an image input layer with 'zerocenter'
    normalization, specify the average image using the AverageImage property.