Elena Ranguelova

Results 51 comments of Elena Ranguelova

Should be fixed in the LIME [text tutorial](https://github.com/dianna-ai/dianna/blob/main/tutorials/explainers/LIME/lime_text.ipynb) and RISE [text tutorial](https://github.com/dianna-ai/dianna/blob/main/tutorials/explainers/RISE/rise_text.ipynb) indeed! Colormap - 'bwr'. Red should indicate a positive contribution towards the decision and blue- a negative contribution.

[like] Elena Ranguelova reacted to your message: ________________________________ From: SarahAlidoost ***@***.***> Sent: Thursday, May 23, 2024 12:16:06 PM To: dianna-ai/dianna ***@***.***> Cc: Elena Ranguelova ***@***.***>; Comment ***@***.***> Subject: Re: [dianna-ai/dianna]...

Also, we are now not using HTML visualization anymore, right? Then we should edit the 4. Visualization section of the notebook to remove that claim.

@SarahAlidoost it's easy for me to fix it within my PR #769 as I just did the same for the overview notebook.

Fixed the colormap issue (though notebooks can be simplified now, by not using the colormap explicitly) with a few commits: https://github.com/dianna-ai/dianna/pull/769/commits/e0d1ee84627fa442e57e4e92d356f76091474f0b; https://github.com/dianna-ai/dianna/pull/769/commits/6e843b16d736a0f69f8cf80da99275624f21609d in https://github.com/dianna-ai/dianna/pull/769/ , but found out dianna breaks...

[like] Elena Ranguelova reacted to your message: ________________________________ From: SarahAlidoost ***@***.***> Sent: Monday, May 27, 2024 11:21:34 AM To: dianna-ai/dianna ***@***.***> Cc: Elena Ranguelova ***@***.***>; Assign ***@***.***> Subject: Re: [dianna-ai/dianna]...

Fixed by commits https://github.com/dianna-ai/dianna/pull/776/commits/643b478ff8a5466982c3ff9826fb3e1ecff2d0df, https://github.com/dianna-ai/dianna/pull/776/commits/c461e7ca316008bda958ee13fa7f26ffd978d799 and https://github.com/dianna-ai/dianna/pull/776/commits/1302a4f1c651bd121f74304b5f4738ae3c41fec1. @loostrum, @SarahAlidoost , @cwmeijer it's very strange that RISE gives always only positive scores even for very negative reviews! See RISE for text...

Perhaps a relevant blog post on mixed data and multiple inputs: https://pyimagesearch.com/2019/02/04/keras-multiple-inputs-and-mixed-data/

Related user stories: > As a domain user I want to see an ML model’s prediction probabilities for classifying an article to a specific genre class and these probabilities for...

@cwmeijer , we need to check if the blocking is solved by solving issue #138.