where do the probabilities come from
play

Where do the probabilities come from? Probabilities come from: - PowerPoint PPT Presentation

Where do the probabilities come from? Probabilities come from: Experts Data D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 1 / 6 Learning probabilities the simplest case Observe tosses of thumbtack:


  1. Where do the probabilities come from? Probabilities come from: ◮ Experts ◮ Data � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 1 / 6

  2. Learning probabilities — the simplest case Observe tosses of thumbtack: Tails Heads n 0 instances of Heads = false n 1 instances of Heads = true what should we use as P ( heads )? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 2 / 6

  3. Learning probabilities — the simplest case Observe tosses of thumbtack: Tails Heads n 0 instances of Heads = false n 1 instances of Heads = true what should we use as P ( heads )? n 1 Empirical frequency: P ( heads ) = n 0 + n 1 � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 2 / 6

  4. Learning probabilities — the simplest case Observe tosses of thumbtack: Tails Heads n 0 instances of Heads = false n 1 instances of Heads = true what should we use as P ( heads )? n 1 Empirical frequency: P ( heads ) = n 0 + n 1 n 1 + 1 Laplace smoothing [1812]: P ( heads ) = n 0 + n 1 + 2 � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 2 / 6

  5. Learning probabilities — the simplest case Observe tosses of thumbtack: Tails Heads n 0 instances of Heads = false n 1 instances of Heads = true what should we use as P ( heads )? n 1 Empirical frequency: P ( heads ) = n 0 + n 1 n 1 + 1 Laplace smoothing [1812]: P ( heads ) = n 0 + n 1 + 2 n 1 + c 1 Informed priors: P ( heads ) = n 0 + n 1 + c 0 + c 1 for some informed pseudo counts c 0 , c 1 > 0. c 0 = 1, c 1 = 1, expressed ignorance (uniform prior) Pseudo-counts convey prior knowledge. Consider: “how much more would I believe α if I had seen one example with α true than if I has seen no examples with α true?” � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 2 / 6

  6. Learning probabilities — the simplest case Observe tosses of thumbtack: Tails Heads n 0 instances of Heads = false n 1 instances of Heads = true what should we use as P ( heads )? n 1 Empirical frequency: P ( heads ) = n 0 + n 1 n 1 + 1 Laplace smoothing [1812]: P ( heads ) = n 0 + n 1 + 2 n 1 + c 1 Informed priors: P ( heads ) = n 0 + n 1 + c 0 + c 1 for some informed pseudo counts c 0 , c 1 > 0. c 0 = 1, c 1 = 1, expressed ignorance (uniform prior) Pseudo-counts convey prior knowledge. Consider: “how much more would I believe α if I had seen one example with α true than if I has seen no examples with α true?” — empirical frequency overfits to the data. � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 2 / 6

  7. Example of Overfitting We have a web site where people rate restaurants with 1 to 5 stars. We want to report the most liked restaurant(s) — the one predicted to have the best future ratings. How can we determine the most liked restaurant? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 3 / 6

  8. Example of Overfitting We have a web site where people rate restaurants with 1 to 5 stars. We want to report the most liked restaurant(s) — the one predicted to have the best future ratings. How can we determine the most liked restaurant? Are the restaurants with the highest average rating the most liked restaurants? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 3 / 6

  9. Example of Overfitting We have a web site where people rate restaurants with 1 to 5 stars. We want to report the most liked restaurant(s) — the one predicted to have the best future ratings. How can we determine the most liked restaurant? Are the restaurants with the highest average rating the most liked restaurants? Which restaurants have the highest average rating? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 3 / 6

  10. Example of Overfitting We have a web site where people rate restaurants with 1 to 5 stars. We want to report the most liked restaurant(s) — the one predicted to have the best future ratings. How can we determine the most liked restaurant? Are the restaurants with the highest average rating the most liked restaurants? Which restaurants have the highest average rating? Which restaurants have a rating of 5? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 3 / 6

  11. Example of Overfitting We have a web site where people rate restaurants with 1 to 5 stars. We want to report the most liked restaurant(s) — the one predicted to have the best future ratings. How can we determine the most liked restaurant? Are the restaurants with the highest average rating the most liked restaurants? Which restaurants have the highest average rating? Which restaurants have a rating of 5? ◮ Only restaurants with few ratings have an average rating of 5. � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 3 / 6

  12. Example of Overfitting We have a web site where people rate restaurants with 1 to 5 stars. We want to report the most liked restaurant(s) — the one predicted to have the best future ratings. How can we determine the most liked restaurant? Are the restaurants with the highest average rating the most liked restaurants? Which restaurants have the highest average rating? Which restaurants have a rating of 5? ◮ Only restaurants with few ratings have an average rating of 5. Solution: add some “average” ratings for each restaurant! � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 3 / 6

  13. Bayesian Learning Probability of Heads … Toss 1 Toss 2 Toss 11 aispace: http://artint.info/code/aispace/beta.xml Probablity of Heads is a random variable representing the probability of heads. Range is { 0 . 0 , 0 . 1 , 0 . 2 , . . . , 0 . 9 , 1 . 0 } or interval [0 , 1]. P ( Toss # n = Heads | Probablity of Heads = v ) = � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 4 / 6

  14. Bayesian Learning Probability of Heads … Toss 1 Toss 2 Toss 11 aispace: http://artint.info/code/aispace/beta.xml Probablity of Heads is a random variable representing the probability of heads. Range is { 0 . 0 , 0 . 1 , 0 . 2 , . . . , 0 . 9 , 1 . 0 } or interval [0 , 1]. P ( Toss # n = Heads | Probablity of Heads = v ) = v Toss # i is independent of Toss # j (for i � = j ) given Probablity of Heads i.i.d. or independent and identically distributed. � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 4 / 6

  15. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

  16. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. What probabilities are required? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

  17. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. What probabilities are required? What counts are required? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

  18. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. What probabilities are required? What counts are required? number of times each help page h i is the best one number of times word w j is used when h i is the help page. � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

  19. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. What probabilities are required? What counts are required? number of times each help page h i is the best one number of times word w j is used when h i is the help page. When can the counts be updated? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

  20. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. What probabilities are required? What counts are required? number of times each help page h i is the best one number of times word w j is used when h i is the help page. When can the counts be updated? When the correct page is found. � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

  21. Naive Bayes Classifier: User’s request for help H "able" "absent" "add" "zoom" . . . H is the help page the user is interested in. We observe the words in the query. What probabilities are required? What counts are required? number of times each help page h i is the best one number of times word w j is used when h i is the help page. When can the counts be updated? When the correct page is found. What prior counts should be used? Can they be zero? � D. Poole and A. Mackworth 2019 c Artificial Intelligence, Lecture 10.1 5 / 6

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend