This Monday and Tuesday (14. – 15. October 2019) I was in Berlin, participating at the #Python and #ML Summit. Actually, #Berlin is a nice city, one of my favourites in Germany, thus if you are not from there, only this reason is good enough go for the Python and ML Summit:
So, the two summits – ML and Python ran in parallel and one could have chosen to which event to go. It started with presentation by the key-note speaker Oliver Zeigermann, who presented 4 case studies, united under the subject “What can man do with ML today”. The case studies included:
- Predicting crash risk for perspective customers
- Germany’s consumption of energy overtime
- Traffic sign recognition
- Deep reinforcement learning
Then I went to “Building a Python Package” by Oz Tiram. Together with him, we built a python package from scratch, following best practices for Python. Some of them are to be found here – https://docs.python.org/3/distributing.
After lunch, I decided to go to the workshop for “Neural Networks with Pytorch”, organized by Chi Nhan Nguyen. It was run on Jupyter and was featuring the framework PyTorch with practical examples. 
At the end of the first day, there was a night quiz, where me and my teammate won our round. Thus, I got a deep learning crash course voucher. I have not started watching it, but considering the fact that the video course is less than 2 hours, I hope to find time soon.
I do not win every competition I participate in, but sometimes it happens🤓 For those wondering, the question is “Which is the first city, which reach 10M inhabitants?” and the chart moves per year, starting from the 15. century😇🙃 https://t.co/i5TxbPVHiq
— vitoshAcademy (@vitoshacademy) October 16, 2019
On the second day, the opening topic was “ML revolution in the public administration”, by André Gode.
On the first talk, I have chosen to “Finding solutions with Machine Learning – Better done than perfect” by Jan Hedtfeld and Lena Müller-Ontjes. The idea was to define a problem based on the characteristics of a given person. The characteristics of the person were discovered through various pictures of their daily life. With a design-thinking-method, a solution was proposed. At the end, the solution was built with Lego blocks. In the case of our group, the solution included an interactive wall in the kindergarten of the user, who was having a problem of not keeping in touch a lot with the world around her and being a bit “offline”:
|
1 2 3 4 5 6 7 8 9 10 11 12 |
conda install -y pip conda install -y -c conda-forge ipython-sql conda install -y -c anaconda beautifulsoup4 conda install -y -c anaconda gensim pip install pyLDAvis conda install -y -c anaconda word2vec conda install -y -c anaconda requests conda install -y -c anaconda nltk conda install -y -c conda-forge spacy python -m spacy download en python -m spacy download de conda install -y -c conda-forge wordcloud |
#MachineLearning analysis puts #CheapHotel and #Brooklyn into same cluster, based on #TripAdvisor sample data at #Python_Summit(#PythonSummit) in #Berlin pic.twitter.com/prR4G1zk4j
— vitoshAcademy (@vitoshacademy) October 15, 2019
It was built upon real data from an accommodation forum, using NLP, Clustering, Topic Modeling, Classification and Universal embedding, presented in real time through Jupyter notebooks. I was actually fascinated by the workshop, and I would dedicate a separate article to its technologies explicitly.
Pretty much this was my 2 day stay in #Berlin, having fun and learning about Machine Learning and Python.
