Machine Learning Zoomcamp Update: Tuesday, 7 September 2021
Date: 7 September 2021
Today, I completed the Sessions 1.1, 1.2 and 1.3.
Session 1.1 - Introduction to Machine Learning
This session gave an introduction to how ML works.
Key takeaways:
- If an expert can do something with reasonalbe accuracy then it is highly probable then we can make a Machine Learning model to achieve similar accuracy on the same task.
-
Features + Target + Machine Learning algorithm -> Model
Features + Model -> Predictions
Session 1.2 - ML vs Rule-Based Systems
This session explains how ML is better than traditional rule-based systems using an example of e-mail spam detection.
Key takeaways:
- It is tough to create a good rule-based system for complex tasks like detecting spam e-mails as rules need to be updated as spammers change their strategy regularly.
- Creating rules and using these rules as features can be a good start for creating a ML model.
Session 1.3 - Supervised Machine Learning
This session gave an introduction to Supervised Machine Learning.
Key takeaways:
- Supervised ML models are like functions which take in features and return target predictions.
- The problem is called a regression problem when the target values are continuous (numerical).
- The problem is called a classification problem when the target values are categorical. When the number of target classes are two it is called binary classification and when the target classes are more than two it is called multiclass classification.
- Ranking problems are those in which we find the probability that the user will like a product/result and then rank them based on these probability before showing them to the user. Example: Google Search results.
Estimated Time Taken: 1 hour 10 minutes