15. Januar 2018 3 min to read

Deep Learning Dive at Data Science Retreat

Category : Artificial Intelligence, Data Science, Deep Learning, Teaching

Namaste! It is me again. Your AI-Guru! I just come back from Berlin and I would love to tell you a story. A story about a Deep Learning Dive. As a yogi and certified yoga teacher, I sometimes go to retreats. A retreat is a place and a time for learning, contemplating and application. There you usually deepen your spiritual understanding and practice. Usually with like-minded individuals. A fact that really fuels your learning. Data Science Retreat in Berlin is the same, but different. It is the place where aspiring Data-Science-yogis meet. And I was happy to teach Deep Learning to lots of beautiful minds as their guru for three days.

Deep Learning is for everyone!

When I arrived, I met an interdisciplinary team of students. And I immediately remembered: Deep Learning is not restricted to Computer Science. As a computer scientist, I have the feeling that more and more fields of research and application are slowly but steadily converging towards IT. And I definitely saw this in the team. It became clear. Deep Learning is a powerful tool for biology, chemistry, linguistics, art and music. It is not the toy of us computer-nerds. Those days are long gone!

Deep Learning Dive at Data Science Retreat. Talking and doing.

My personal teaching style is special and different in some ways. I am happy to admit: I hate everything that is built around Powerpoint presentations. And I do not like talking to people for hours on end without any interaction. And this is exactly the point. For me teaching and learning is highly interactive. My approach is based on two principles. Inspiring discussions and hands-on practice.

There is never the one and only way to teaching a thing. Never. There are always many. And the best one usually comes out of the current situation. Out of the moment. You cannot plan this. This is why I interleave theory and practice. I explain some principles and after that ensure that the ideas are manifested in experiments immediately.

What did the students take home? Around ten excellent Jupyter notebooks with condensed and clear knowledge. And lots of ideas and good best-practices in their heads and our notebooks. During three days of training we averaged around three notebooks per day, which is quite something. Three days… What a joyride!

Three days: Deep intro, stabilizing knowledge, broadening the horizon.

Well, on the first day, we arrived at our topic at hand. A little theory. What is Deep Learning and why is it so accessible right now? How come that basically everyone has the resources to train Neural Networks from scratch? Why is it an illusion that Neural Networks are black magic? Those are examples of questions that we answered early on.

It did not take us long to start applying our fresh and new knowledge. Immediately after the introduction, we trained our first Neural Network. That was right before the lunch break. It is that easy.

The second day was all about solidifying knowledge. Cementing what we have learned on the first day. We slowly turned our fear of overfitting into a mission towards overfitting. Because once you overfit with a high training accuracy you can boost validation accuracy with a lots of strategies. Run into overfitting and then compensate it! And finally on that day, before crossing the finishing line, we opened our minds to Convolutional Neural Networks. Those powerful Neural Networks for image processing.

And the third and final day was all about broadening our knowledge to the maximum. We put the CNNs heavily into practice. And then turned towards Natural Language Processing. After an full spectrum wrap-up we said goodbye. Three days of Deep Learning Dive were over.

What is the core learning of our Deep Learning Dive?

Definitely our mantra: It is time-to-market. It is true: You can learn Deep Learning in no time. And three days is no time. As a joke I said once: Go to the Apple Store, buy a Mac, go home and get started. After about an hour you will be training your first Neural Network. And this could not be closer to the truth. It is a hard fact. I love these times!

But that was just three days… From here on out, it will be all about applying and getting more experiences. I wish the participants all the best! It has been a pleasure working with you! You are on the right track! And I learned a lot from you. Thanks!

Stay in touch.

I hope you liked the article. Why not stay in touch? You will find me at LinkedIn, XING and Facebook. Please add me if you like and feel free to like, comment and share my humble contributions to the world of AI. Thank you! I am looking forward to talking to you!

Leave a comment