First steps: Deep Learning

We’re done with week VIII: Reinforcement Learning and our first steps in Deep Learning.

 

Neurons

Of course, we spent a lot of time on learning about the concept of Neural Networks, before we started writing code. The instructors explain everything very well so that we have a good basic understanding of the components of an ANN and how they work together. After a good night of sleep I will spend my Sunday reading about gradient descent optimization algorithms and listening to Geoffrey Hinton. Here’s what happened yesterday:

 

I’m the King of the World!

So I got my first ANN working, I followed the instructions without understanding everything fully, but fortunately (or unfortunately) everything is so well prepared in the course that it went smoothly and I got good results.

We were super excited. And since I was not satisfied with my Kaggle Titanic results, I got even more excited and thought wow, this is the way I will boost the accuracy and rank up easily.

So I started working with the Titanic data and building an ANN model, but I wasn’t really sure what exact parameters to input, the hidden layers, the number of epochs, .. So I googled and tried things out, but wasn’t able to achieve higher accuracy than 84%. When I submitted the results, I was ranked the same as with previous Classification models. After 11 hours of studying with my group members, I was tired and gave up.

 

A new day

A Taifun is coming to Tokyo today. Perfect timing. I gave up yesterday, but today is a new day and I’m determined to solve the Titanic problem, rank up and join the top 10%. I’ll keep you posted.

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s