The Author Online Book Forums are Moving

The Author Online Book Forums will soon redirect to Manning's liveBook and liveVideo. All book forum content will migrate to liveBook's discussion forum and all video forum content will migrate to liveVideo. Log in to liveBook or liveVideo with your Manning credentials to join the discussion!

Thank you for your engagement in the AoF over the years! We look forward to offering you a more enhanced forum experience.

293244 (1) [Avatar] Offline
#1
Hi,

Thanks for the book on Machine Learning with TensorFlow. The clarification section on "When to use a metric other than L2 norm in real world" - Chapter 1, was not exactly very clear. Could you elaborate some more on that part ?

thanks
Nishant Shukla (52) [Avatar] Offline
#2
Sure! I rephrased it a little to make it easier to understand. You'll see it in the latest book update.
Let’s say you’re working for a new search-engine start-up trying to compete with Google.
Your boss assigns you the task of using machine learning to personalize the search results for each user.
A good goal might be that users shouldn’t see 5 or more incorrect search-results per month.
A year’s worth of user-data is a 12-dimension vector (where each month of the year is a dimension), indicating number of incorrect results shown per month.
You are trying to satisfy the condition that the L-infinity norm of this vector must below 5.
Suppose instead that your boss changes his/her mind and requires that less than 5 erroneous search-results are allowed for the entire year.
In this case, you are trying to achieve a L1 norm below 5.
Actually, your boss changes his or her mind again.
Now, the number of months with erroneous search-results should be less than 5.
In that case, you are trying to achieve an L0 norm below 5.