Eureqa solution expression

redenton
Linear Actuator

Eureqa solution expression

I set "EUREQA_target_expression_string" to the following:


lgrhoeq = f0(cos(thYear-f1()),cos(thMlt-f2()) L,F,F3,AE,Kp,Kp3,Dst,Bysw,Bzsw,Vsw,Npsw,Pdsw)


The idea is that thYear and thMlt are angles, and I want to have phase angle offsets, f1() and f2(). But Eureqa is telling me that there is an error in the format. Does anybody know why this format is not okay? Here's the log:

 

DataRobot Blueprint Console
[12-07-2021 21:48:02] Cross Validation # 1 started
[12-07-2021 21:50:07] [parse_error] Error parsing expression while Parsing expression: Parse failure at offset 1
[12-07-2021 21:50:07] Expression: lgrhoeq = f0(cos(thYear-f1()),cos(thMlt-f2()),L,F,F3,AE,Kp,Kp3,Dst,Bysw,Bzsw,Vsw,Npsw,Pdsw)
[12-07-2021 21:50:07] ^- Parse error here
[12-07-2021 21:50:07] Variable names:
[12-07-2021 21:50:07] <row>
[12-07-2021 21:50:07] thYear
[12-07-2021 21:50:07] L
[12-07-2021 21:50:07] thMlt
[12-07-2021 21:50:07] F
[12-07-2021 21:50:07] F3
[12-07-2021 21:50:07] AE
[12-07-2021 21:50:07] Kp
[12-07-2021 21:50:07] Kp3
[12-07-2021 21:50:07] Dst
[12-07-2021 21:50:07] Bysw
[12-07-2021 21:50:07] Bzsw
[12-07-2021 21:50:07] Vsw
[12-07-2021 21:50:07] Npsw
[12-07-2021 21:50:07] Pdsw
[12-07-2021 21:50:07] Target
[12-07-2021 21:50:07] 
[12-07-2021 21:50:07] (1) All solutions in Eureqa modeler had infinite fitness scores.
[12-07-2021 21:50:07] Cross Validation # 1 failed
[12-07-2021 21:50:11] Cross Validation # 1 failed
1 Solution

Accepted Solutions

Hi Redenton, I think you need to replace the target feature name "lgrhoeq" with the word "Target". I believe the blueprint always overrides the target name to "Target". There's more information on the target expression formatting here: 

https://app.datarobot.com/model-docs/tasks/EQC-Eureqa-Classifier.html

View solution in original post

8 Replies

Hey @redenton 

Not a Eureqa expert here but are you missing a ',' between 'cos(thMlt-f2()) L'
lgrhoeq = f0(cos(thYear-f1()),cos(thMlt-f2()) L,F,F3,AE,Kp,Kp3,Dst,Bysw,Bzsw,Vsw,Npsw,Pdsw)

Should it be:

lgrhoeq = f0(cos(thYear-f1()),cos(thMlt-f2()),L,F,F3,AE,Kp,Kp3,Dst,Bysw,Bzsw,Vsw,Npsw,Pdsw) 

f0 should not be included, nor all those external processing terms. You’re defining the function- not the formula 

@IraWatt Ira, thank you very much for getting back to me! Unfortunately, that doesn't seem to fix it:

 

DataRobot Blueprint Console
[12-09-2021 18:53:50] Cross Validation # 1 started
[12-09-2021 18:55:49] Error parsing expression while Parsing expression: Parse failure at offset 1
[12-09-2021 18:55:49] Expression: lgrhoeq = f0(cos(thYear-f1()),cos(thMlt-f2()),L,F,F3,AE,Kp,Kp3,Dst,Bysw,Bzsw,Vsw,Npsw,Pdsw)
[12-09-2021 18:55:49] ^- Parse error here
[12-09-2021 18:55:49] Variable names:
[12-09-2021 18:55:49] <row>
[12-09-2021 18:55:49] thYear
[12-09-2021 18:55:49] L
[12-09-2021 18:55:49] thMlt
[12-09-2021 18:55:49] F
[12-09-2021 18:55:49] F3
[12-09-2021 18:55:49] AE
[12-09-2021 18:55:49] Kp
[12-09-2021 18:55:49] Kp3
[12-09-2021 18:55:49] Dst
[12-09-2021 18:55:49] Bysw
[12-09-2021 18:55:49] Bzsw
[12-09-2021 18:55:49] Vsw
[12-09-2021 18:55:49] Npsw
[12-09-2021 18:55:49] Pdsw
[12-09-2021 18:55:49] Target
[12-09-2021 18:55:49] 
[12-09-2021 18:55:49] (1) All solutions in Eureqa modeler had infinite fitness scores.
[12-09-2021 18:55:49] Cross Validation # 1 failed
[12-09-2021 18:55:53] Cross Validation # 1 failed
0 Kudos

@Ryan 

Ryan, thank you for getting back to me! How do you suggest that I write it? The point is to specify that I want to of the terms to be of the form cos(th - f1()). I was looking here,

https://docs.datarobot.com/en/docs/modeling/reference/eureqa-ref/custom-expressions.html,

and I see examples like this:

Target = f(x, w, x^f1(), w^f2())

Thanks, Richard

0 Kudos

Hi Redenton, I think you need to replace the target feature name "lgrhoeq" with the word "Target". I believe the blueprint always overrides the target name to "Target". There's more information on the target expression formatting here: 

https://app.datarobot.com/model-docs/tasks/EQC-Eureqa-Classifier.html

@MichaelS

Thank you very much for responding. But I'm still getting a parse error. Here's the log file:
DataRobot Blueprint Console
[01-06-2022 21:00:16] Cross Validation # 1 started
[01-06-2022 21:00:26] [parse_error] Error parsing expression while Parsing expression: Parse failure at offset 16
[01-06-2022 21:00:26] Expression: Target = f(cos(thYear-f1()),cos(thMlt-f2()),L,F,F3,AE,Kp,Kp3,Dst,Bysw,Bzsw,Vsw,Npsw,Pdsw)
[01-06-2022 21:00:26] ^- Parse error here
[01-06-2022 21:00:26] Variable names:
[01-06-2022 21:00:26] <row>
[01-06-2022 21:00:26] cyear
[01-06-2022 21:00:26] syear
[01-06-2022 21:00:26] L
[01-06-2022 21:00:26] cmlt
[01-06-2022 21:00:26] smlt
[01-06-2022 21:00:26] F
[01-06-2022 21:00:26] F3
[01-06-2022 21:00:26] AE
[01-06-2022 21:00:26] Kp
[01-06-2022 21:00:26] Kp3
[01-06-2022 21:00:26] Dst
[01-06-2022 21:00:26] Bysw
[01-06-2022 21:00:26] Bzsw
[01-06-2022 21:00:26] Vsw
[01-06-2022 21:00:26] Npsw
[01-06-2022 21:00:26] Pdsw
[01-06-2022 21:00:26] Target
[01-06-2022 21:00:26]
[01-06-2022 21:00:26] (1) All solutions in Eureqa modeler had infinite fitness scores.
[01-06-2022 21:00:26] Cross Validation # 1 failed
[01-06-2022 21:00:30] Cross Validation # 1 failed

0 Kudos

@redenton It looks like some of the variables used in the target expression aren't in the list of available variables names in the error message. It looks like maybe "thYear" and "thMlt" might have been renamed or not in the dataset?

 

The available features comes from the feature list used to train the model. By default it uses "informative features" which is all features except those with low information (e.g. all missing values). Maybe double check this if it they were filtered out by mistake. 

0 Kudos

@MichaelS You're right. I'm embarrassed to say that I uploaded the wrong data file with the wrong arguments. So, I will try to make sure in the future I have the right data file loaded, and will use "Target =" in the expression. Thank you.