Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

34
And What You Can Take from Each Pedro Domingos University of Washington

Transcript of Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Page 1: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

And What You Can Take from Each

Pedro Domingos University of Washington

Page 2: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Evolution Experience

Culture

Page 3: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Evolution Experience

Culture Computers

Page 4: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Most of the knowledge in the world in the future is going to be extracted by machines and will reside in machines. – Yann LeCun, Director of AI Research, Facebook

Page 5: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

1. Fill in gaps in existing knowledge

2. Emulate the brain

3. Simulate evolution

4. Systematically reduce uncertainty

5. Notice similarities between old and new

Page 6: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Tribe Origins Master Algorithm

Symbolists Logic, philosophy Inverse deduction

Connectionists Neuroscience Backpropagation

Evolutionaries Evolutionary biology Genetic programming

Bayesians Statistics Probabilistic inference

Analogizers Psychology Kernel machines

Page 7: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Tom Mitchell Steve Muggleton Ross Quinlan

Page 8: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Addition Subtraction

2 + 2 ――― = ? ――

2 + ? ――― = 4 ――

Page 9: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Deduction

Socrates is human

+ Humans are mortal . ―――――――――――

= ?

Induction

Socrates is human

+ ? ――――――――――― = Socrates is mortal

―――――――――― ――――――――――

Page 10: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 11: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Yann LeCun Geoff Hinton Yoshua Bengio

Page 12: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 13: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 14: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 15: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 16: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

John Koza John Holland

Hod Lipson

Page 17: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 18: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 19: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 20: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

David Heckerman Judea Pearl Michael Jordan

Page 21: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 22: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 23: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 24: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Peter Hart Vladimir Vapnik Douglas Hofstadter

Page 25: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 26: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 27: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15
Page 28: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Tribe Problem Solution

Symbolists Knowledge composition Inverse deduction

Connectionists Credit assignment Backpropagation

Evolutionaries Structure discovery Genetic programming

Bayesians Uncertainty Probabilistic inference

Analogizers Similarity Kernel machines

Page 29: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Tribe Problem Solution

Symbolists Knowledge composition Inverse deduction

Connectionists Credit assignment Backpropagation

Evolutionaries Structure discovery Genetic programming

Bayesians Uncertainty Probabilistic inference

Analogizers Similarity Kernel machines

But what we really need is a single algorithm that solves all five!

Page 30: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

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

Page 31: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Much remains to be done . . .

We need your ideas

Page 32: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

Home Robots

Cancer Cures 360o Recommenders

World Wide Brains

Page 33: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

If we used all our technology resources, we could actually give people personalized recommendations for every step of your life. – Aneesh Chopra, former CTO of the U.S.

Page 34: Pedro Domingos, Professor, University of Washington at MLconf ATL - 9/18/15

MLconf NYC Pek Lum

MLconf NYC Pek Lum

MLconf NYC Edo Liberty

MLconf NYC Edo Liberty

Raji Balasuubramaniyan, Senior Data Scientist, Manheim at MLconf ATL - 9/18/15

Raji Balasuubramaniyan, Senior Data Scientist, Manheim at MLconf ATL - 9/18/15

David Talby, SVP Engineering, Atigeo at MLconf ATL - 9/18/15

David Talby, SVP Engineering, Atigeo at MLconf ATL - 9/18/15

Music recommendations @ MLConf 2014

Music recommendations @ MLConf 2014

Sandy Ryza – Software Engineer, Cloudera at MLconf ATL

Sandy Ryza – Software Engineer, Cloudera at MLconf ATL

MLconf NYC Animashree Anandkumar

MLconf NYC Animashree Anandkumar

MLconf NYC Chang Wang

MLconf NYC Chang Wang

Funda Gunes, Senior Research Statistician Developer & Patrick Koch, Principal Data Scientist, SAS Institute Inc. at MLconf ATL 2016

Funda Gunes, Senior Research Statistician Developer & Patrick Koch, Principal Data Scientist, SAS Institute Inc. at MLconf ATL 2016

Kaz Sato, Evangelist, Google at MLconf ATL 2016

Kaz Sato, Evangelist, Google at MLconf ATL 2016

Ted Dunning, Chief Application Architect, MapR at MLconf ATL - 9/18/15

Ted Dunning, Chief Application Architect, MapR at MLconf ATL - 9/18/15

Le Song, Assistant Professor, College of Computing, Georgia Institute of Technology at MLconf ATL 2016

Le Song, Assistant Professor, College of Computing, Georgia Institute of Technology at MLconf ATL 2016

Jorge A. Castañón, Data Scientist, IBM at MLconf ATL - 9/18/15

Jorge A. Castañón, Data Scientist, IBM at MLconf ATL - 9/18/15

Talwalkar mlconf (1)

Talwalkar mlconf (1)

Chris Fregly, Research Scientist, PipelineIO at MLconf ATL 2016

Chris Fregly, Research Scientist, PipelineIO at MLconf ATL 2016

Talha Obaid, Email Security, Symantec at MLconf ATL 2017

Talha Obaid, Email Security, Symantec at MLconf ATL 2017

Jake Mannix, MLconf 2013

Jake Mannix, MLconf 2013

Hussein Mehanna, Engineering Director, ML Core - Facebook at MLconf ATL 2016

Hussein Mehanna, Engineering Director, ML Core - Facebook at MLconf ATL 2016

Elizabeth Elhassani – Director, Enterprise Analytics & Insights, LexisNexis at MLconf ATL

Elizabeth Elhassani – Director, Enterprise Analytics & Insights, LexisNexis at MLconf ATL

Mahoney mlconf-nov13

Mahoney mlconf-nov13