How to learn a Machine Learning algorithm?
Everything you need to consider while approaching to learn a #MachineLearning algorithm ๐
A thread ๐งต
Everything you need to consider while approaching to learn a #MachineLearning algorithm ๐
A thread ๐งต
1. Get the intuition behind the algorithm (i.e its core ideas and why the algorithm is there in the first place).
---
2. Get the mathematical intuition behind the algorithm (understand the math working under the hood).
---
---
2. Get the mathematical intuition behind the algorithm (understand the math working under the hood).
---
3. For what the algorithm is used (regression/classification/both) and how it is modified to fit different scenarios.
---
4. How the algorithm works with numerical and categorical data?
---
---
4. How the algorithm works with numerical and categorical data?
---
5. How the algorithm works with different types of data (tabular, text, etc).
---
6. Impact of outliers and missing data on the algorithm.
---
---
6. Impact of outliers and missing data on the algorithm.
---
7. Is feature scaling required or not and why?
---
8. Impact of the algorithm on imbalanced data and how to fix that?
---
---
8. Impact of the algorithm on imbalanced data and how to fix that?
---
9. Bias-Variance Trade-off for the algorithm.
---
10. Assumptions of the algorithm.
---
---
10. Assumptions of the algorithm.
---
11. Where it can be used and where not? (this point is kind-of combination of the above-mentioned points)
---
12. Understand the Time-Space complexity of it.
---
13. Implement it with scikit-learn and read the docs (pretty well-explained docs).
---
12. Understand the Time-Space complexity of it.
---
13. Implement it with scikit-learn and read the docs (pretty well-explained docs).
14. Get a brief idea of what each parameter in the scikit-learn function does (again, read it from the docs).
---
15. Extensions of the algorithm (if any).
---
16. Advantages and Disadvantages of it.
---
15. Extensions of the algorithm (if any).
---
16. Advantages and Disadvantages of it.
17. Interpretability of the algorithm.
---
18. Implement it from scratch (optional, but highly recommended).
---
18. Implement it from scratch (optional, but highly recommended).
Loading suggestions...