| Line 4: | Line 4: | ||
| − | In | + | In Lecture 5, we observed that Bayes rule minimizes the probability of error. One student pointed out an example that raises doubts about the validity of this statement. This gave rise to an interesting [[ECE662_topic2_discussions|online discussion]]. In particular, a student [[EE662Sp10OptimalPrediction|proved that the example previously proposed performs worse]] than following Bayes rule. |
| + | We extended the discussion of Bayes rule to the case of continuous-valued feature vectors, including a discussion of the expected loss (called "risk") when following Bayes rule. | ||
Previous: [[Lecture4ECE662S10|Lecture 4]] | Previous: [[Lecture4ECE662S10|Lecture 4]] | ||
| − | |||
Next: [[Lecture6ECE662S10|Lecture 6]] | Next: [[Lecture6ECE662S10|Lecture 6]] | ||
---- | ---- | ||
Revision as of 07:17, 12 April 2010
Details of Lecture 5, ECE662 Spring 2010
In Lecture 5, we observed that Bayes rule minimizes the probability of error. One student pointed out an example that raises doubts about the validity of this statement. This gave rise to an interesting online discussion. In particular, a student proved that the example previously proposed performs worse than following Bayes rule.
We extended the discussion of Bayes rule to the case of continuous-valued feature vectors, including a discussion of the expected loss (called "risk") when following Bayes rule.
Previous: Lecture 4 Next: Lecture 6
