07. Validation

Validation

---

Topic: Validation

Course: GMLC

Date: 18 February 2019 

Professor: Not specified

---

Resources

---

• https://developers.google.com/machine-learning/crash-course/validation/check-your-intuition

• https://developers.google.com/machine-learning/crash-course/validation/video-lecture

• https://developers.google.com/machine-learning/crash-course/validation/another-partition

• https://developers.google.com/machine-learning/crash-course/validation/programming-exercise

Key Points

---

• 2 subsets my cause overfitting on the test set if we are tweaking the model after confirming training set on the test set

• Validation set

  • a slice in partition used to evaluate training mode on and then tweak hyperparameters based on validation set’s result

  • Used to prevent overfitting on test set

  • Prevent exposure on the test set

• Test set - used to confirm results and notice overfitting on the validation set

Check your understanding

---

• What are the 3 partitions we split the initial Dataset into?

• What is the importance of Validation set or why is it dangerous to repeat iterations on test set?

Summary of Notes

---

• 3 partitions in a dataset is a common workflow because we can tweak the hyperparameters after evaluating the trained data on a validation set rather than exposing the final test set and cause overfitting on it.