11. Logistic Regression

Logistic Regression

---

Topic: Logistic Regression

Course: GMLC

Date: 25 February 2019 

Professor: Not specified

---

Resources

---

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

• https://developers.google.com/machine-learning/crash-course/logistic-regression/calculating-a-probability

• https://developers.google.com/machine-learning/crash-course/logistic-regression/model-training

Key Points

---

• Logistic regression

  • Model which uses the sigmoid function to a linear prediction, thus generating a probability for each possible discrete label value in classification problems

  • y=1/(1+e−z)

  • Y - the output of the logistic regression model

  • Z - b1 + w1x2 + w2x2...

  • Returned probability of logistic regression can either be used as is  or using binary classification (spam or not spam)

  • Generates probability

• Log loss

  • Loss function for logistic regression

  • Minimising the loss function yields a maximum likelihood estimate

  • Regularization

    • L2

    • Early stopping

    • L1

Check your understanding

---

• What is logistic regression used for?

• What loss function does logistic regression use?

Summary of Notes

---

• Logistic regression is widely practiced in machine learning as a provider of the probability of a label

• Log loss is a loss function used for logistic regression