what do noisy datapoints tell us about the true signal
play

What do noisy datapoints tell us about the true signal? C. R. Hogg - PowerPoint PPT Presentation

Sep 03, 2023 •358 likes •1.89k views

Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What do noisy datapoints tell us about the true signal? C. R. Hogg November 16, 2011 1/31

  1. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What is Bayesian Analysis? Essence of Bayes: • 2 questions for every guess (i.e. every θ ) 1. How likely does it make the actual data ? “LIKELIHOOD” 2. How plausible is it? “PRIOR” • Combine them to answer the main question : 1. What is your new probability, now that you’ve seen the data? “POSTERIOR” (Rev. Thomas Bayes, c. 1701 – 1761) 6/31

  2. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What is Bayesian Analysis? Essence of Bayes: • 2 questions for every guess (i.e. every θ ) 1. How likely does it make the actual data ? “LIKELIHOOD” 2. How plausible is it? “PRIOR” • Combine them to answer the main question : 1. What is your new probability, now that you’ve seen the data? “POSTERIOR” (Rev. Thomas Bayes, c. 1701 – 1761) 6/31

  3. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Likelihood of function p ( y | f ) • Example: artificial dataset 40 • Noise model: Poisson Intensity (number of counts) p ( y | f ) = f y e − f y ! 20 • Assume independent pixels 0 0.0 0.5 1.0 Scattering vector Q (1/A) 7/31

  4. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Likelihood of function p ( y | f ) • Example: artificial dataset 40 • Noise model: Poisson Intensity (number of counts) p ( y | f ) = f y e − f y ! 20 • Assume independent pixels 0 0.0 0.5 1.0 Scattering vector Q (1/A) 7/31

  5. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Likelihood of function p ( y | f ) • Example: artificial dataset 40 • Noise model: Poisson Intensity (number of counts) p ( y | f ) = f y e − f y ! 20 • Assume independent pixels 0 0.0 0.5 1.0 Scattering vector Q (1/A) 7/31

  6. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Likelihood of function p ( y | f ) • Example: artificial dataset 40 • Noise model: Poisson Intensity (number of counts) p ( y | f ) = f y e − f y ! 20 • Assume independent pixels 0 0.0 0.5 1.0 Scattering vector Q (1/A) 7/31

  7. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Likelihood of function p ( y | f ) • Example: artificial dataset 40 • Noise model: Poisson Intensity (number of counts) p ( y | f ) = f y e − f y ! 20 • Assume independent pixels • Problem: not plausible • (What makes a function 0 “plausible”?) 0.0 0.5 1.0 Scattering vector Q (1/A) 7/31

  8. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) • Assume smooth and continuous • No functional form assumed 8/31

  9. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  10. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  11. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  12. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  13. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  14. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  15. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Plausibility” of function p ( f ) 40 Intensity (number of counts) • Assume smooth and continuous • No functional form 20 assumed • Naturally: unrelated to data 0 0.0 0.5 1.0 Scattering vector Q (1/A) 8/31

  16. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) 40 • “Best of both worlds”: Intensity (number of counts) a Plausible curves, which b fit the data 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 9/31

  17. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) 40 • “Best of both worlds”: Intensity (number of counts) a Plausible curves, which b fit the data 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 9/31

  18. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) 40 • “Best of both worlds”: Intensity (number of counts) a Plausible curves, which b fit the data 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 9/31

  19. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) 40 • “Best of both worlds”: Intensity (number of counts) a Plausible curves, which b fit the data 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 9/31

  20. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) 40 • “Best of both worlds”: Intensity (number of counts) a Plausible curves, which b fit the data 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 9/31

  21. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) 40 • “Best of both worlds”: Intensity (number of counts) a Plausible curves, which b fit the data 20 • To represent uncertainty: show many guesses • (Or, summarize them. . . ) 0 0.0 0.5 1.0 Scattering vector Q (1/A) 9/31

  22. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) Quantitative uncertainty visuals Noisy Data 40 35 Intensity (no. of counts) 30 25 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 Q (1/A) 9/31

  23. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posterior probability p ( f | y ) Quantitative uncertainty visuals Noisy Data 40 True Curve! 35 Intensity (no. of counts) 30 25 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 Q (1/A) 9/31

  24. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recap: Bayesian denoising Plausible curves 40 Intensity (number of counts) 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 10/31

  25. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recap: Bayesian denoising Curves which fit the data 40 Intensity (number of counts) 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 10/31

  26. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recap: Bayesian denoising Plausible curves which fit the data 40 Intensity (number of counts) 20 0 0.0 0.5 1.0 Scattering vector Q (1/A) 10/31

  27. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity • calculate probabilities for its values 11/31

  28. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , • calculate probabilities for its values • take “random draws” (roll the die, flip the coin. . . ) 11/31

  29. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , • calculate probabilities for its values • take “random draws” (roll the die, flip the coin. . . ) 11/31

  30. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , • calculate probabilities for its values • take “random draws” (roll the die, flip the coin. . . ) 11/31

  31. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , • calculate probabilities for its values • take “random draws” (roll the die, flip the coin. . . ) 11/31

  32. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values • take “random draws” (roll the die, flip the coin. . . ) 11/31

  33. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , • take “random draws” (roll the die, flip the coin. . . ) 11/31

  34. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , • take “random draws” (roll the die, flip the coin. . . ) 11/31

  35. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , • take “random draws” (roll the die, flip the coin. . . ) 11/31

  36. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) 11/31

  37. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  38. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  39. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  40. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  41. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  42. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  43. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  44. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Random variables; Random functions 1 1 1 1 1 p = 1 6 6 6 6 6 6 • Random variable F : an uncertain quantity F = , , , , , • calculate probabilities for its values , , , , . . . • take “random draws” (roll the die, flip the coin. . . ) • Random function F ( x ) ? F(x) 11/31 x

  45. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values F(x) 12/31

  46. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values F(x) 12/31

  47. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  48. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  49. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  50. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  51. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  52. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  53. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  54. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation F(x) 12/31

  55. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation • co rrelation × variance : covariance F(x) 12/31

  56. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to think about “random functions”? • Function: a collection of individual values • Every value is a random variable, with. . . 1. variance 2. correlation • co rrelation × variance : covariance • Gaussian Process: F(x) • Every point is a Random Variable • Any (finite) subset has Gaussian joint distribution 12/31

  57. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix • How to read the matrix? 0 Cov 2500 100 1 0 2 -100 0 3 0 1 2 3 0 1 2 3 X X 13/31

  58. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix • How to read the matrix? 1. By individual entries 0 Cov 2500 100 1 0 2 -100 0 3 0 1 2 3 0 1 2 3 X X 13/31

  59. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix • How to read the matrix? 1. By individual entries 0 Cov 2500 100 1 0 2 -100 0 3 0 1 2 3 0 1 2 3 X X 13/31

  60. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix • How to read the matrix? 1. By individual entries 0 Cov 2500 100 1 0 2 -100 0 3 0 1 2 3 0 1 2 3 X X 13/31

  61. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix • How to read the matrix? 1. By individual entries 0 Cov 2500 2. As a whole 100 1 (central stripe) 0 • Intensity: 2 height of features -100 0 3 • Width: 0 1 2 3 0 1 2 3 X X width of features 13/31

  62. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix 0 Stripe Intensity = Feature Height 100 1 0 2 • How to read the matrix? -100 3 0 1 2 3 0 1 2 3 X X 1. By individual entries 0 Cov 2500 2. As a whole 100 1 (central stripe) 0 • Intensity: 2 height of features -100 0 3 • Width: 0 1 2 3 0 1 2 3 X X width of features 13/31

  63. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How to read a Covariance Matrix 0 Stripe Intensity = Feature Height 100 1 0 2 • How to read the matrix? -100 3 0 1 2 3 0 1 2 3 X X 1. By individual entries 0 Cov 2500 2. As a whole 100 1 (central stripe) 0 • Intensity: 2 height of features -100 0 3 • Width: 0 1 2 3 0 1 2 3 X X width of features 0 Stripe Width = Feature Width 100 1 0 2 -100 3 0 1 2 3 0 1 2 3 X X 13/31

  64. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Example 1: Hydrocarbon combustion (Dave Sheen and Wing Tsang, NIST, Div. 632) Hydrocarbon burning simulations • Need (many!) reaction rate constants • Measured individually • Predictions are precise, quantitative 14/31

  65. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Example 1: Hydrocarbon combustion (Dave Sheen and Wing Tsang, NIST, Div. 632) Hydrocarbon burning simulations • Need (many!) reaction rate constants • Measured individually • Predictions are precise, quantitative, wrong 14/31

  66. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrocarbon combustion: flame speed experiments • Datapoints (from several experiments) Flame speed 40 30 Speed (cm/s) 20 10 0 0.6 0.8 1.0 1.2 1.4 1.6 Fuel-to-oxygen ratio 15/31

  67. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrocarbon combustion: flame speed experiments • Datapoints (from several experiments) • Model : lengthscales Flame speed ℓ and σ f 40 30 Speed (cm/s) 20 10 0 0.6 0.8 1.0 1.2 1.4 1.6 Fuel-to-oxygen ratio 15/31

  68. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrocarbon combustion: flame speed experiments • Datapoints (from several experiments) • Model : lengthscales Flame speed ℓ and σ f 40 • ± 1 σ range 30 Speed (cm/s) 20 10 0 0.6 0.8 1.0 1.2 1.4 1.6 Fuel-to-oxygen ratio 15/31

  69. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrocarbon combustion: flame speed experiments • Datapoints (from several experiments) • Model : lengthscales Flame speed ℓ and σ f 40 • ± 1 σ range 30 Speed (cm/s) • See also: individual curves 20 10 0 0.6 0.8 1.0 1.2 1.4 1.6 Fuel-to-oxygen ratio 15/31

  70. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrocarbon combustion: flame speed experiments • Datapoints (from several experiments) • Model : lengthscales Flame speed ℓ and σ f 40 • ± 1 σ range 30 Speed (cm/s) • See also: individual curves 20 • But where did this model come from. . . ? 10 0 0.6 0.8 1.0 1.2 1.4 1.6 Fuel-to-oxygen ratio 15/31

  71. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor, intro • William of Occam c. 1288 - c. 1348 • Gave us Occam’s Razor 16/31

  72. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor, intro • William of Occam c. 1288 - c. 1348 • Gave us Occam’s Razor • (slightly paraphrased in the name of science) 16/31

  73. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor, intro • William of Occam c. 1288 - c. 1348 • Gave us Occam’s Razor • (slightly paraphrased in the name of science) • Claim: use probability , get this automatically • (And, quantitative , too!) 16/31

  74. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor, intro • William of Occam c. 1288 - c. 1348 • Gave us Occam’s Razor MODEL 1: • (slightly paraphrased in flat the name of science) • Claim: use probability , get this automatically • (And, quantitative , too!) • Example: 3 models. . . 16/31

  75. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor, intro • William of Occam c. 1288 - c. 1348 • Gave us Occam’s Razor MODEL 2: • (slightly paraphrased in flat + line the name of science) • Claim: use probability , get this automatically • (And, quantitative , too!) • Example: 3 models. . . 16/31

  76. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor, intro • William of Occam c. 1288 - c. 1348 • Gave us Occam’s Razor MODEL 3: • (slightly paraphrased in flat + line + wiggle the name of science) • Claim: use probability , get this automatically • (And, quantitative , too!) • Example: 3 models. . . 16/31

  77. Overview Bayesian Analysis Gaussian Processes Scattering Curves Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Occam’s razor in action • Some models can explain more datasets 17/31

Recommend

Noisy Channel Coding: Correlated Random Variables & Communication over a Noisy Channel Toni

Noisy Channel Coding: Correlated Random Variables & Communication over a Noisy Channel Toni

Noisy Channel Coding: Correlated Random Variables & Communication over a Noisy Channel Toni Hirvonen Helsinki University of Technology Laboratory of Acoustics and Audio Signal Processing Toni.Hirvonen@hut.fi T-61.182 Special Course in

242 views • 20 slides
Kalman filter Kalman Filter Kalman filter is used to filter true system states from noisy

Kalman filter Kalman Filter Kalman filter is used to filter true system states from noisy

Kalman filter Kalman Filter Kalman filter is used to filter true system states from noisy measurements T o apply the filter, we have to have good model of a system Common use cases: localization of robots, spaceships, ... radar

500 views • 11 slides
Sanntidsdeling av data i en API- lst verden What is event driven? Wait Signal += (event)

Sanntidsdeling av data i en API- lst verden What is event driven? Wait Signal += (event)

Sanntidsdeling av data i en API- lst verden What is event driven? Wait Signal += (event) => you say; Business TimeSeries / Discreet Truth of the system DataPoints Named intentionally Measurements Represents the domain

668 views • 29 slides
Adaptive Signal Recovery by Convex Optimization Dmitrii Ostrovskii CWI, Amsterdam 19 April 2018

Adaptive Signal Recovery by Convex Optimization Dmitrii Ostrovskii CWI, Amsterdam 19 April 2018

Adaptive Signal Recovery by Convex Optimization Dmitrii Ostrovskii CWI, Amsterdam 19 April 2018 Signal denoising problem Recover complex signal x = ( x ) , = n , ..., n , from noisy observations y = x + , = n ,

697 views • 43 slides
Practical denoising of clipped or overexposed noisy images Alessandro Foi www.cs.tut.fi/~foi

Practical denoising of clipped or overexposed noisy images Alessandro Foi www.cs.tut.fi/~foi

1 TAMPERE UNIVERSITY OF TECHNOLOGY Department of Signal Processing Transforms and Spectral Methods Group Practical denoising of clipped or overexposed noisy images Alessandro Foi www.cs.tut.fi/~foi EUSIPCO-2008 - Lausanne, Switzerland,

2.88k views • 40 slides
Speech Recognition and Synthesis Dan Klein UC Berkeley Noisy Channel Model: ASR We want to

Speech Recognition and Synthesis Dan Klein UC Berkeley Noisy Channel Model: ASR We want to

Language Models Speech Recognition and Synthesis Dan Klein UC Berkeley Noisy Channel Model: ASR We want to predict a sentence given acoustics: The noisy-channel approach: The Speech Signal Acoustic model: score fit Language model: score

414 views • 20 slides
TRUE IT A LI A N T A STE PROJECT WH A T IS THE TRUE IT A LI A N T A STE PROJECT ? The True

TRUE IT A LI A N T A STE PROJECT WH A T IS THE TRUE IT A LI A N T A STE PROJECT ? The True

I T A L Y - A M E R I C A C H A M B E R O F C O M M E R C E TRUE IT A LI A N T A STE PROJECT WH A T IS THE TRUE IT A LI A N T A STE PROJECT ? The True Italian Taste project is a large-scale educational and promotional initiative aimed at

80 views • 6 slides
Unsupervised Learning About this class Build a model for your data. Which datapoints

Unsupervised Learning About this class Build a model for your data. Which datapoints

Unsupervised Learning About this class Build a model for your data. Which datapoints Unsupervised learning are similar? k -means Clustering Nowadays lots of work on using unlabeled data to improve the performance of supervised learn-

450 views • 6 slides
Personal Statements TRUE FALSE TRUE FALSE TRUE There is a 4,000 character

Personal Statements TRUE FALSE TRUE FALSE TRUE There is a 4,000 character

Personal Statements TRUE FALSE TRUE FALSE TRUE There is a 4,000 character limit (including spaces) or 47 lines. It comes to about 1.5 A4 pages. TRUE FALSE The university will receive your whole UCAS application which

443 views • 43 slides
Graphical Exploratory Analysis Using Take a fixed collection of datapoints : ( x 1 , y 1 ) , ( x 2

Graphical Exploratory Analysis Using Take a fixed collection of datapoints : ( x 1 , y 1 ) , ( x 2

Bivariate halfspace depth (Tukey depth) Graphical Exploratory Analysis Using Take a fixed collection of datapoints : ( x 1 , y 1 ) , ( x 2 , y 2 ) , . . . , ( x n , y n ) . Halfspace Depth Given an arbitrary point ( x , y ) : take all (closed)

208 views • 5 slides
between the mountains and the sea True Majesty. True North. True Scotland CruiseAberdeenshire

between the mountains and the sea True Majesty. True North. True Scotland CruiseAberdeenshire

between the mountains and the sea True Majesty. True North. True Scotland CruiseAberdeenshire offers: CruiseA eAber berdeensh deenshir ire is a working alliance between Aberdeen A joined-up Port and Destination offering Harbour

590 views • 32 slides
Multi-parameter regularization for ill-posed problems with noisy right hand side and noisy

Multi-parameter regularization for ill-posed problems with noisy right hand side and noisy

Introduction TLS, RTLS and DRTLS Multi-parameter reg. Error bounds Computational aspects Conclusions Multi-parameter regularization for ill-posed problems with noisy right hand side and noisy operator Ulrich Tautenhahn University of

231 views • 21 slides
CSE217 INTRODUCTION TO DATA SCIENCE LECTURE 6: LEARNING PRINCIPLES Spring 2019 Marion Neumann

CSE217 INTRODUCTION TO DATA SCIENCE LECTURE 6: LEARNING PRINCIPLES Spring 2019 Marion Neumann

CSE217 INTRODUCTION TO DATA SCIENCE LECTURE 6: LEARNING PRINCIPLES Spring 2019 Marion Neumann RECAP: MACHINE LEARNING Workflow 2 NOISE noisy samples from true function 3 WHY IS NOISE A PROBLEM? small random sample from the noisy

747 views • 19 slides
Waveform Generation Fundamental part of signal processing is the signal. Within the

Waveform Generation Fundamental part of signal processing is the signal. Within the

Waveform Generation Fundamental part of signal processing is the signal. Within the context of hardware implementations of signal processing this signal may arise from: A local waverform/signal generator. Continuous stream from a

492 views • 26 slides
Review of DSP 1 Signal and Systems: Signal are represented mathematically as functions of

Review of DSP 1 Signal and Systems: Signal are represented mathematically as functions of

Review of DSP 1 Signal and Systems: Signal are represented mathematically as functions of one or more independent variables. Digital signal processing deals with the transformation of signal that are discrete in both amplitude and

456 views • 34 slides
LEARNING IN GAMES WITH NOISY PAYOFF OBSERVATIONS Background and motivation Preliminaries The

LEARNING IN GAMES WITH NOISY PAYOFF OBSERVATIONS Background and motivation Preliminaries The

LEARNING IN GAMES WITH NOISY PAYOFF OBSERVATIONS Background and motivation Preliminaries The core scheme Learning with noisy feedback Mario Bravo Panayotis Mertikopoulos Universidad de Santiago de Chile CNRS Laboratoire

841 views • 65 slides
OCR Errors by Michael Barz Motivation In general: How to get information out of noisy input?

OCR Errors by Michael Barz Motivation In general: How to get information out of noisy input?

OCR Errors by Michael Barz Motivation In general: How to get information out of noisy input? Dealing with noisy input (scan/fax/e- mail) in written form Approach: Combination of diverse NLP tools in one pipeline Optical

543 views • 29 slides
A Random Walk through CS70 bounds for computing whether you have an even number of 1s as true?

A Random Walk through CS70 bounds for computing whether you have an even number of 1s as true?

A Random Walk through CS70 bounds for computing whether you have an even number of 1s as true? Then Q true, then R true. If P not true, then both sides reduce CS70 Summer 2016 - Lecture 8B 4 Applications: Circuits Boolean logic encompasses a

336 views • 5 slides
Sampling a Signal an analog signal together with some samples of the signal. The samples

Sampling a Signal an analog signal together with some samples of the signal. The samples

Sampling a Signal an analog signal together with some samples of the signal. The samples shown are equally spaced and simply pick off the value of the underlying analog value of the underlying analog signal at the appropriate times. If

412 views • 23 slides
signal electron is single mono energetic delayed The signal electron

signal electron is single mono energetic delayed The signal electron

Proton - Production high-P low-P BG e - e - Signal - Spectrometer Magnet -e Conversion Muonic Atom Formation in-flight Production Secondary Beamline Target signal electron is

166 views • 16 slides
Unsupervised, Fast and Precise Recognition of Digital Arcs in Noisy Images T. P . NGUYEN, B.

Unsupervised, Fast and Precise Recognition of Digital Arcs in Noisy Images T. P . NGUYEN, B.

Introduction Arc segmentation Unsupervised Noise Detection A framework for arc recognition along noisy curves Experimentations Conclusions and futur work Unsupervised, Fast and Precise Recognition of Digital Arcs in Noisy Images T. P .

550 views • 51 slides
# of true positives true positive rate = # of known positives (Proportion of actual positives

# of true positives true positive rate = # of known positives (Proportion of actual positives

True positive rate (Sensitivity) # of true positives true positive rate = # of known positives (Proportion of actual positives that are correctly identified) True negative rate (Specificity) # of true negatives true negative rate = # of known

402 views • 18 slides
# of true positives true positive rate = # of known positives (Proportion of actual positives

# of true positives true positive rate = # of known positives (Proportion of actual positives

True positive rate (Sensitivity) # of true positives true positive rate = # of known positives (Proportion of actual positives that are correctly identified) True negative rate (Specificity) # of true negatives true negative rate = # of known

475 views • 24 slides
Formal Modeling in Cognitive Science 1 Noisy Channel Model Channel Capacity Lecture 29: Noisy

Formal Modeling in Cognitive Science 1 Noisy Channel Model Channel Capacity Lecture 29: Noisy

Noisy Channel Model Noisy Channel Model Kullback-Leibler Divergence Kullback-Leibler Divergence Cross-entropy Cross-entropy Formal Modeling in Cognitive Science 1 Noisy Channel Model Channel Capacity Lecture 29: Noisy Channel Model and

236 views • 5 slides

More recommend