Number to Number LSTM RNN

I want to make a numeric to numeric prediction (of time series data) using LSTM RNN. I only have 1 week of online courses in machine learning. I’m not sure but I don’t think you have a video covering this case and the AutoML wizard says it doesn’t cover number to number.

I am getting the following error: “TypeError : can only concatenate tuple (not “dict”) to tuple”

Following the architecture in my course, I am attempting it with Input (none, 1) then 4 LSTM layers of (none, 50) then Dense layer of (none, 1) then Output.

Can you share screenshot of the data tab?

Also if you can share the model config, that would be great. To export the model, click export on the export button on model tab.

Now that I have entered an input dimension for the Dense layer, I get a different error message:

SyntaxError : keyword argument repeated (, line 1)

I attached a screenshot of the data tab, as requested.

Here is the model config:

data:
dataset: {name: train2, samples: 34727, type: private}
datasetLoadOption: batch
kfold: 1
mapping:
‘-0.540600852’:
options: {Normalization: true, Scaling: 1}
port: InputPort0
shape: ‘’
type: Numeric
numPorts: 1
samples: {split: 1, test: 0, training: 27781, validation: 6945}
shuffle: false
model:
connections:

  • {source: LSTM_3, target: LSTM_4}
  • {source: LSTM_2, target: LSTM_3}
  • {source: LSTM_4, target: Dense_1}
  • {source: Dense_1, target: Output_1}
  • {source: LSTM_1, target: LSTM_2}
  • {source: Input_1, target: LSTM_1}
    layers:
  • args: {}
    class: Input
    name: Input_1
    x: 542
    y: 15
  • args: {}
    class: Output
    name: Output_1
    x: 560
    y: 634
  • args: {input_dim: ‘50’, output_dim: ‘1’}
    class: Dense
    name: Dense_1
    x: 558
    y: 536
  • args: {output_dim: ‘50’}
    class: LSTM
    name: LSTM_1
    x: 541
    y: 142
  • args: {output_dim: ‘50’}
    class: LSTM
    name: LSTM_2
    x: 550
    y: 244
  • args: {output_dim: ‘50’}
    class: LSTM
    name: LSTM_3
    x: 549
    y: 346
  • args: {output_dim: ‘50’}
    class: LSTM
    name: LSTM_4
    x: 553
    y: 440
    params:
    batch_size: 32
    loss_func: categorical_crossentropy
    num_epoch: 10
    optimizer: {name: Adadelta}
    project: test 2

Thanks for your help!