Deciding what to Try Next

• Should be expected by now average expert of machine learning
• Know truly what algorithm for each problem

• What to do when the result is largely different from the actual output
• Prioritize small subset of feature, more impotance ones
• lambda(regularization parameters)
• Unfortunately, most common are choosing by instinct, not by science
• Choosing which avenues by these simple method
• These techniques will part the correct way from the rest of the avenues
• Next two videos, evaluate learning algorithm, and next discuss about this technique

• Insight from the test of which promising way to try
• Diagnostic can be time consuming, but still worthed compared to spend many months only to find out that the avenues that we choose is wrong

• Next try few things to improve machine learning system

• Previous post
• Next post