please add model correlation heatmap

[aplonska]

No description provided.

[pplonski]

Spearman correlation is computed between all trained models at the end of AutoML training. The example below:

For classification the correlation is computed based on proba values. For multiclass the average of correlation for each class is done.

[pplonski]

Please resize in the case of many models.

[pplonski]

I added larger figure in the case of the many models:

[Tonywhitemin]

Hi pplonski,
Good day! I am now study this heatmap but have some questions on it...
Could you please help?
If follow the above commend you said:
"For classification the correlation is computed based on proba values. For multiclass the average of correlation for each class is done."
Since this heatmap is bases on different models, may I know it used what parameters or value to do the correlation?
Could you tell me what is the mechanism of it?
Thanks!

[pplonski]

The correlation is computed between each model's predictions and target. It is computed with scipy.stats.spearmanr.

[Tonywhitemin]

Thank you so much pplonski!