Presentation is loading. Please wait.

Presentation is loading. Please wait.

Pedro Domingos University of Washington. Traditional Programming Machine Learning Computer Data Algorithm Output Computer Data Output Algorithm.

Similar presentations


Presentation on theme: "Pedro Domingos University of Washington. Traditional Programming Machine Learning Computer Data Algorithm Output Computer Data Output Algorithm."— Presentation transcript:

1 Pedro Domingos University of Washington

2

3 Traditional Programming Machine Learning Computer Data Algorithm Output Computer Data Output Algorithm

4 Traditional Programming Machine Learning Computer Data Algorithm Output Master Algorithm Data Output Algorithm

5 TribeOriginsMaster Algorithm SymbolistsLogic, philosophyInverse deduction ConnectionistsNeuroscienceBackpropagation EvolutionariesEvolutionary biologyGenetic programming BayesiansStatisticsProbabilistic inference AnalogizersPsychologyKernel machines

6 Tom MitchellSteve MuggletonRoss Quinlan

7 AdditionSubtraction 2 + 2 ――― = ? ―― 2 + ? ――― = 4 ――

8 Deduction Socrates is human + Humans are mortal. ――――――――――― = ? Induction Socrates is human + ? ――――――――――― = Socrates is mortal ――――――――――

9

10 Yann LeCunGeoff HintonYoshua Bengio

11

12

13

14

15 John Koza John HollandHod Lipson

16

17

18

19 David HeckermanJudea PearlMichael Jordan

20

21

22

23 Peter HartVladimir VapnikDouglas Hofstadter

24

25

26

27 TribeProblemSolution SymbolistsKnowledge compositionInverse deduction ConnectionistsCredit assignmentBackpropagation EvolutionariesStructure discoveryGenetic programming BayesiansUncertaintyProbabilistic inference AnalogizersSimilarityKernel machines

28 TribeProblemSolution SymbolistsKnowledge compositionInverse deduction ConnectionistsCredit assignmentBackpropagation EvolutionariesStructure discoveryGenetic programming BayesiansUncertaintyProbabilistic inference AnalogizersSimilarityKernel machines But what we really need is a single algorithm that solves all five!

29 Representation Probabilistic logic (e.g., Markov logic networks) Weighted formulas → Distribution over states Evaluation Posterior probability User-defined objective function Optimization Formula discovery: Genetic programming Weight learning: Backpropagation

30 Much remains to be done... We need your ideas

31 Home Robots Cancer Cures360 o Recommenders World Wide Brains

32


Download ppt "Pedro Domingos University of Washington. Traditional Programming Machine Learning Computer Data Algorithm Output Computer Data Output Algorithm."

Similar presentations


Ads by Google