Hi Everyone,
I am working on machine learning and artificial intelligence for quite a few years. Over the years, I developed/trained so many black box models. Overall, the models seem to work but what they lack is explainability like that of linear regression models. Howerver, I am thinking to build explainable (may be not 100%) models so that we can increase the trust of the end user and make small adjustments to it. My question is whether something like that is possible? Any references will be appreciated.
We were concerned with a similar question and reviewed current approaches for interpreting and visualizing deep neural networks. Our research resulted in a classification scheme and a comprehensive overview over the current state of the art as presented throughout the paper:
Conference Paper Demystifying the Black Box: A Classification Scheme for Inte...
Please let us know if you find our work helpful for your own research.
Kind regards
Patrick
There is this well written book, where you can find more information about this topic: https://christophm.github.io/interpretable-ml-book/
Came here to share same book as Philipp Probst did above ^^. I believe it is good place to start. Also see the O'Reilly book attached.
I would take a look at this link: https://www.google.com/search?q=explaining+machine+learning+models&rlz=1C1CHBF_enUS847US847&oq=explining+machine+learni&aqs=chrome.2.69i57j0l5.29533j1j4&sourceid=chrome&ie=UTF-8
Best, D. Booth
- Badges
- Science topic