A Year of LearningI have been struggling with gaining proper control over machine learning techniques. I've read online articles, reproduced walkthrough results, even done some online coursework, but I've continued to struggle when adapting these techniques to new problems.
I feel a little tiny bit hopeless.
This isn't a particularly unfamiliar feeling for me -- this is the great monster to be overcome whenever learning something new and challenging. It's just a feeling to remind me that it's worth keeping going, to prove once again that I've still got it. This time around, I reckon it's going to take a little longer than normal.
For that reason I've decided that I need a specific schedule, a set of clear objectives and something at stake. Let's work backwards. There's little objectively at stake here. I'm not really a gambling man, I don't need a new job, and this is really a personal exercise. What's really at stake is just my own sense of achievement. I though I'd try to raise those stakes through this blog, by committing my progress into form and making it public.
Let's talk about the objectives. In no particular order...
- Be able to competently apply deep learning techniques to new image-based problems
- Give a presentation at PyCon AU 2015 on the topic of deep learning in Python
- Write a blog post per week on this topic for a year, or until feel I have achieve the other objectives on the list
- Enter a kaggle competition with a result in the top 50%
- Enter a kaggle competition with a result in the top 25%
- Enter a kaggle competition with a result in the top 10%
- Write up all my code, results and learnings into this blog post for the benefit of others
- Publish my code into a public repository and have at least one other person actually make use of it
I'd really like to hear from anyone who might be reading about this. Is there anything in particular I can expand on? What about this story is interesting for you?