Data Scientist, Economist, Design Thinker @Fraunhofer IMW in Leipzig, Germany
Leipzig, Deutschland
Joined August 2012
- Tweets12
- Following64
- Followers14
- Likes32
Pinned Tweet
How to find industry partners for public research? In our new #IEEE paper @yuricampbll, Friedrich Dornbusch, Anna Pohle and I present the classification approach behind our tool Corporate-Match, which helps 70 @Fraunhofer institutes with that. Paper: ieeexplore.ieee.org/document… 1/5
Karl Trela retweeted
📢We are excited to announce our paper "
Deep Learning-based Computational Job Market Analysis: A Survey on Skill Extraction and Classification from Job Postings" has been accepted to #EACL2024 at the @nlp4hr workshop.
In cooperation with @NLPnorth and @MaiNLPlab.
What a great result of one summer seminar with students: an entire book. Wish my seminars had been that interesting and productive back then.
What are some limitations of interpretable machine learning methods?
This summer, our students worked on this question. We compiled the results in a free online book, which we release today. 🎉🎉🎉
Find out more:
compstat-lmu.github.io/iml_m…
I think every Data Scientist has noticed at some point how important and sometimes difficult it is to make your #MachineLearning model interpetable. This looks like an awesome resource with state of the art ways to do it. Will surely check it out. Thanks @ChristophMolnar!
2 years, 250 pages, 1,219 commits, and 78,480 words:
I am very proud to say that today I published the 1st edition of "Interpretable Machine Learning". 🎉🎉🎉
Web: christophm.github.io/interpr…
Leanpub: leanpub.com/interpretable-ma…
1
Turns out the combination of econometric data, geographic company-institute distance and WZ-industry-codes (German equivalent to NACE) performs the best. Texts which describe the industry branches instead of codes perform slightly worse on average, but with less variance. 4/5
1
1
We actually use a variant of this classification approach for our Corporate-Match prototype and our colleagues at @Fraunhofer are quite satisfied with the results. We are still working on improvements and would love to apply it to different organizations in future. 5/5
1
I really enjoyed discussing our poster at #mlprague together with @algorithm87, on how we try find the right industry partners for our research institutes @Fraunhofer using #MachineLearning methods like text-classification and hybrid recommender systems. Thanks for dropping by!
Interesting future challenge for machine learning put forward by MITs Tomaso Poggio: " to learn like children"...with just a few labeled observations, possibly wit a lot of unlabeled data. Children do not need a million of pictures to learn how a car looks like. #mlprague
1
4
Wow, what a way to start a conference with playing this organ! #mlprague
1