What does mean the convergence of algorithm? Please discuss.

hi @MunchCrunch19 can you please share your process?

@sgenzer Well the model applied in azure machine learning Studio , 

Hello @MunchCrunch19

If you are asking about the convergence of an ML algorithm, then the convergence is when the algorithm function will stay in a set error range even though you iterate it several times.

In a simple statement, when a model converges there won't be a significant reduction in model error.

I think in your case, as you are using random search of hyperparameters, your model will iterate for multiple sets of parameters and at some point, it will converge and you won't see much improvement in your model after that converging point.

@varunm1 Well, this question asked by a reviewer as I submitted my paper in the journal, I don't know how to respond to him/her! Please, I need your input on this, 
I used stratified 10 Fold Cross-validation for the mentioned model and hyperparameters tunning, I used 17 Iteration for each hyperparameter tuning. In the end, I cross-validate the model, Let me show you the hyperparameters results and the Quantile loss for 0.07 Quantile and 0.95 quantile and Average quantile, which I got for each iteration. Please see the Picture attached 

Kindly explain the mentioned algorithm convergence in this regard!

Is this the exact question he/she asked?

Can you provide his/her statement? Did he/she ask you to prove convergence?

From an algorithm point of view, I don't have much knowledge about the quantile regression algorithm and I need to take a look at it. 

Did you use a stratified 10 fold inside random parameter search? Are you using parameter search on the training data side of cross-validation or on the whole data (some researchers does this)?

@varunm1 The exact question he/she asked      " Please organize the modeling approach in Section 4 (Methodology) as a modeling or identification algorithm, with clear steps. What about the convergence of this algorithm? Please discuss."


I have used stratified 10FOLD cross-validation using hyperparameter tuning with random search, after getting the best parameter values through Hyperparameter (Random search) I then used these values in the model and used 10 fold cross-validation and cross-validate the model. 

Please Kindly see the process in the Below photos with hyperparameters tuning and without hyperparameter tuning. 

@varunm1          Waiting for your input in this regard!

hi @MunchCrunch19 those screenshots do not look like RapidMiner....