Machine Learning | Bayesian Linear Regression
This video starts with the inspiration behind the Bayesian Linear Regression and compares it to the Frequentist (OLS, Lasso, Ridge) regressions to highlight real-life conditions where Bayesian regression is better than the other popular linear regression models.
The video then goes into the details of Bayesian mathematics and lays a strong foundation of understanding of the role of Weight distribution and how as per Bayesian assumptions the maximum a-posterior estimate of W is computed instead of the maximum likelihood ones.
The video then also graphically demonstrates how it solver the problem of missing data, noise and generates confidence bands of predictions at each point, which are in turn based on Bayesian probabilities for p(y | w, X) and p(w | y, X).
Links of related videos:
ML | Naive Bayes Classifier:
• Machine Learning | Naive Bayes Classifier ...
ML | Regularization:
• Machine Learning | Regularization - Lasso,...
ML | Bias Variance Trade-Off:
• Machine Learning | Bias Variance Trade-Off