Lessons from Machine Learning: Sleep to Get Ahead
Everyday there are new articles that we need to read on how to be smarter, more productive and make better decisions.
As I write this at 10am, I’ve already counted 13 new articles on Lifehacker promising me the secrets to a better life. I should be excited about these new nuggets of life lessons, instead I feel overwhelmed. Even with dozen of aggregators (Nuzzel, Newsletters) I use to surface the most important articles, it doesn’t help. Trying to construct the prefect algorithm to optimize your life daily, is tiring. A break is much needed but I never take it because of the fear that I’ll be left behind.
Trying to construct the prefect algorithm to optimize your life daily is tiring.
But what if the fear was wrong, what if a break is really what you need to optimize your life and give you enough information to know where to focus on.
The thought that came to me ironically when I was listening to the book Master Algorithm. In the book they describe that the gold standard to testing the effectiveness of an algorithm is to compare it’s performance on a test data set vs. a training data set. When the accuracy is lower, it implies that the algorithm is overfitted and often time what is needed is to generalize the algorithm not complicate it.
What we can learn from this process is that when we construct new algorithms in our lives based on our training data aka past experience, we need to give it time to collect test data. Without test data we can’t determine the effectiveness of the algorithm and we often overfit our algorithm for the current situation making it useless when a variable in our life changes (which is always).
Algorithmic performance on test data is the gold standard to determining how we can get better. Now I’ve given you an excuse to rest, hope you enjoy sleeping in. :)