Google cloud skill boost

---

Main URL

https://www.cloudskillsboost.google/catalog?keywords=&locale=&solution[]=any&role[]=ml---developer&skill-badge[]=any&format[]=any&level[]=any&duration[]=any&language[]=any

---

How Google Does Machine Learning

https://www.cloudskillsboost.google/course_templates/10

Notes:

Backward-looking data = look at historical data ML = make predictive decision base on data

Google translate:

• e.g. take photo of a sign to translate
• 1 model to find the sign
• 1 to read the sign
• OCR
• 1 to detect the language
• 1 to translate the sign
• 1 to superimposed translated text
• maybe 1 to select the font style now using

Google search :

• few years ago logic:
  • a bunch of hand-coded rules, e.g. your location + query text => which result to show
• NOW:
  • DL
  • RankBrain - a DL network for search ranking
  • WHen you click on a link, it may tell google: this result is what you wanted from the query text

ML can solve: anything you are writing rules today

throw away the heuristics just as soon as you have enough data about user preferences

To frame a ML problem

• As an ML problem:
  • what is being predicted
  • what data do we need
• As a software problem:
  • What does the API look like
  • what is the input & output
  • Who will use this service
  • how others are doing this today
• As a data problem:
  • what kind of data we need to collect
  • What to analyze
  • How to react to the prediction

Comments