Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Version spaces

393 views

Published on

Published in: Technology, Education
  • Be the first to comment

  • Be the first to like this

Version spaces

  1. 1. Machine Learning Version Spaces Learning
  2. 2. Introduction: <ul><ul><li>Neural Net approaches </li></ul></ul><ul><ul><li>Symbolic approaches: </li></ul></ul><ul><ul><ul><li>version spaces </li></ul></ul></ul><ul><ul><ul><li>decision trees </li></ul></ul></ul><ul><ul><ul><li>knowledge discovery </li></ul></ul></ul><ul><ul><ul><li>data mining </li></ul></ul></ul><ul><ul><ul><li>speed up learning </li></ul></ul></ul><ul><ul><ul><li>inductive learning </li></ul></ul></ul><ul><ul><ul><li>… </li></ul></ul></ul><ul><li>Very many approaches to machine learning: </li></ul>
  3. 3. Version Spaces A concept learning technique based on refining models of the world.
  4. 4. Concept Learning <ul><li>Example: </li></ul><ul><ul><li>a student has the following observations about having an allergic reaction after meals: </li></ul></ul>Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No + - + - - <ul><ul><li>Concept to learn : under which circumstances do I get an allergic reaction after meals ?? </li></ul></ul>
  5. 5. In general <ul><li>There is a set of all possible events: </li></ul><ul><ul><li>Example: </li></ul></ul><ul><li>There is a boolean function (implicitly) defined on this set. </li></ul><ul><ul><li>Example: </li></ul></ul><ul><li>We have the value of this function for SOME examples only. </li></ul>Restaurant Meal Day Cost 3 X 3 X 7 X 2 = 126 Reaction: Restaurant X Meal X Day X Cost --> Bool Find an inductive ‘guess’ of the concept, that covers all the examples !
  6. 6. Pictured: Set of all possible events                                   Given some examples: + + + + - - - - Find a concept that covers all the positive examples and none of the negative ones!
  7. 7. Non-determinism <ul><li>Many different ways to solve this! </li></ul><ul><li>How to choose ? </li></ul>Set of all possible events                                   + + + + - - - -
  8. 8. An obvious, but bad choice: <ul><li>The concept IS: </li></ul><ul><ul><li>Alma 3 and breakfast and Friday and cheap OR </li></ul></ul><ul><ul><li>Alma 3 and lunch and Saturday and cheap </li></ul></ul><ul><li>Does NOT generalize the examples any !! </li></ul>Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No - - - + +
  9. 9. Pictured: <ul><li>Only the positive examples are positive ! </li></ul>Set of all possible events + + + + - - - -                                  
  10. 10. Equally bad is: <ul><li>The concept is anything EXCEPT : </li></ul><ul><ul><li>De Moete and lunch and Friday and expensive AND </li></ul></ul><ul><ul><li>Sedes and breakfast and Sunday and cheap AND </li></ul></ul><ul><ul><li>Alma 3 and breakfast and Sunday and expensive </li></ul></ul>Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No + - + - -
  11. 11. Pictured: <ul><li>Everything except the negative examples are positive: </li></ul>Set of all possible events                                   + + + + - - - -
  12. 12. Solution: fix a language of hypotheses: <ul><li>We introduce a fix language of concept descriptions. </li></ul>= hypothesis space <ul><li>The concept can only be identified as being one of the hypotheses in this language </li></ul><ul><ul><li>avoids the problem of having ‘useless’ conclusions </li></ul></ul><ul><ul><li>forces some generalization/induction to cover more than just the given examples. </li></ul></ul>
  13. 13. Reaction - Example: Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No + - + - - <ul><li>Every hypothesis is a 4-tuple: </li></ul><ul><ul><li>maximally specific: ex.: [ Sedes, lunch, Monday, cheap] </li></ul></ul><ul><ul><li>combinations of ? and values are allowed: ex.: [ De Moete, ?, ?, expensive] </li></ul></ul><ul><ul><li>or [ ?, lunch, ?, ?] </li></ul></ul><ul><li>One more hypothesis:  (bottom = denotes no example) </li></ul><ul><ul><li>most general hypothesis: [ ?, ?, ?, ?] </li></ul></ul>
  14. 14. Hypotheses relate to sets of possible events       Events    x2      x1 x1 = < Alma 3, lunch, Monday, expensive> x2 = < Sedes, lunch, Sunday, cheap> h1 = [?, lunch, Monday, ?] h2 = [?, lunch, ?, cheap] h3 = [?, lunch, ?, ?] General Specific                 Hypotheses h2 h3 h1
  15. 15. Expressive power of this hypothesis language: <ul><li>Conjunctions of explicit, individual properties </li></ul><ul><li>Examples: </li></ul><ul><ul><li>[?, lunch, ?, cheap] : Meal = lunch  Cost = cheap </li></ul></ul><ul><ul><li>[?, lunch, ?, ?] : Meal = lunch </li></ul></ul><ul><li>In addition to the 2 special hypotheses: </li></ul><ul><ul><li> : nothing </li></ul></ul><ul><ul><li>[?, lunch, Monday, ?] : Meal = lunch  Day = Monday </li></ul></ul><ul><ul><li>[ ?, ?, ?, ?] : anything </li></ul></ul>
  16. 16. Other languages of hypotheses are allowed <ul><li>Example: identify the color of a given sequence of colored objects. </li></ul><ul><li>A useful language of hypotheses: </li></ul><ul><ul><li>the examples: </li></ul></ul><ul><ul><ul><li>red : + </li></ul></ul></ul><ul><ul><ul><li>purple : - </li></ul></ul></ul><ul><ul><ul><li>blue : + </li></ul></ul></ul>any_color mono_color polyo_color red blue green orange purple 
  17. 17. Important about hypothesis languages: <ul><li>They should have a specific <-> general ordering. Corresponding to the set-inclusion of the events they cover. </li></ul>
  18. 18. Defining Concept Learning: <ul><li>Given: </li></ul><ul><ul><li>A set X of possible events : </li></ul></ul><ul><ul><ul><li>Ex.: Eat-events: <Restaurant, Meal, Day, Cost> </li></ul></ul></ul><ul><ul><li>An (unknown) target function c: X -> {-, +} </li></ul></ul><ul><ul><ul><li>Ex.: Reaction: Eat-events -> { -, +} </li></ul></ul></ul><ul><ul><li>A language of hypotheses H </li></ul></ul><ul><ul><ul><li>Ex.: conjunctions: [ ?, lunch, Monday, ?] </li></ul></ul></ul><ul><ul><li>A set of training examples D , with their value under c </li></ul></ul><ul><ul><ul><li>Ex.: (<Alma 3, breakfast,Friday,cheap>,+) , … </li></ul></ul></ul><ul><li>Find: </li></ul><ul><ul><li>A hypothesis h in H such that for all x in D : x is covered by h  c( x ) = + </li></ul></ul>
  19. 19. The inductive learning hypothesis: If a hypothesis approximates the target function well over a sufficiently large number of examples , then the hypothesis will also approximate the target function well on other unobserved examples .
  20. 20. Find-S: a naïve algorithm Replace h by a minimal generalization of h that covers x Return h Initialize : h :=  For each positive training example x in D Do: If h does not cover x :
  21. 21. Reaction example: minimal generalizations = the individual events Generalization = replace something by ? no more positive examples: return h Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No + - + - - Initially: h =  Example 1: h = [Alma 3,breakfast,Friday,cheap]  Example 2: h = [Alma 3, ? , ? ,cheap] 
  22. 22. Properties of Find-S <ul><li>Non-deterministic : </li></ul><ul><ul><li>Depending on H, there maybe several minimal generalizations: </li></ul></ul>Beth Jo Alex Hacker Scientist Footballplayer  H Beth can be minimally generalized in 2 ways to include a new example Jo.
  23. 23. Properties of Find-S (2) <ul><li>May pick incorrect hypothesis (w.r.t. the negative examples): </li></ul>D : Beth Alex Jo + - + Beth Jo Alex Hacker Scientist Footballplayer  H Is a wrong solution
  24. 24. Properties of Find-S (3) <ul><li>Cannot detect inconsistency of the training data: </li></ul><ul><li>Nor inability of the language H to learn the concept: </li></ul>D : Beth Jo Beth + + - D : Beth Alex Jo + - + Jo Alex Scientist  H Beth
  25. 25. Nice about Find-S: <ul><li>It doesn’t have to remember previous examples ! </li></ul>If h already covered the 20 first examples, then h’ will as well h        X H <ul><ul><li>If the previous h already covered all previous examples, then a minimal generalization h’ will too ! </li></ul></ul> h’   
  26. 26. Dual Find-S: Initialize : h := [ ?, ?, .., ?] For each negative training example x in D Do: If h does cover x : Replace h by a minimal specialization of h that does not cover x Return h
  27. 27. Reaction example: Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No + - + - - Initially: h = [ ?, ?, ?, ?] Example 1:  h= [ ?, breakfast, ?, ?] Example 2:  h= [ Alma 3, breakfast, ?, ?] Example 3:  h= [ Alma 3, breakfast, ?, cheap]
  28. 28. Version Spaces: the idea: <ul><li>BUT: do NOT select 1 minimal generalization or specialization at each step, but keep track of ALL minimal generalizations or specializations </li></ul><ul><li>Perform both Find-S and Dual Find-S: </li></ul><ul><ul><li>Find-S deals with positive examples </li></ul></ul><ul><ul><li>Dual Find-S deals with negative examples </li></ul></ul>
  29. 29. Version spaces: initialization <ul><li>The version spaces G and S are initialized to be the smallest and the largest hypotheses only. </li></ul>G = { [ ?, ?, …, ?] } S = {  }
  30. 30. Negative examples: <ul><li>Replace the top hypothesis by ALL minimal specializations that DO NOT cover the negative example. </li></ul>Invariant : only the hypotheses more specific than the ones of G are still possible: they don’t cover the negative example S = {  } G = { [ ?, ?, …, ?] } G = {h1, h2, …, hn}
  31. 31. Positive examples: <ul><li>Replace the bottom hypothesis by ALL minimal generalizations that DO cover the positive example. </li></ul>Invariant : only the hypotheses more general than the ones of S are still possible: they do cover the positive example G = {h1, h2, …, hn} S = {  } S = { h1’, h2’, …,hm’ }
  32. 32. Later: negative examples <ul><li>Replace the all hypotheses in G that cover a next negative example by ALL minimal specializations that DO NOT cover the negative example. </li></ul>Invariant : only the hypotheses more specific than the ones of G are still possible: they don’t cover the negative example S = { h1’, h2’, …,hm’ } G = {h1, h2, …, hn} G = {h1, h21, h22, …, hn}
  33. 33. Later: positive examples <ul><li>Replace the all hypotheses in S that do not cover a next positive example by ALL minimal generalizations that DO cover the example. </li></ul>Invariant : only the hypotheses more general than the ones of S are still possible: they do cover the positive example G = {h1, h21, h22, …, hn} S = { h1’, h2’, …,hm’ } S = { h11’, h12’, h13’, h2’, …,hm’ }
  34. 34. Optimization: negative: <ul><li>Only consider specializations of elements in G that are still more general than some specific hypothesis (in S ) </li></ul>… are more general than ... G = {h1, h21, …, hn} S = { h1’, h2’, …,hm’ } Invariant : on S used !
  35. 35. Optimization: positive: <ul><li>Only consider generalizations of elements in S that are still more specific than some general hypothesis (in G ) </li></ul>… are more specific than ... Invariant : on G used ! G = {h1, h21, h22, …, hn} S = { h13’, h2’, …,hm’ }
  36. 36. Pruning: negative examples <ul><li>The new negative example can also be used to prune all the S - hypotheses that cover the negative example. </li></ul>G = {h1, h21, …, hn} S = { h1’, h3’, …,hm’ } Invariant only works for the previous examples, not the last one Cover the last negative example!
  37. 37. Pruning: positive examples <ul><li>The new positive example can also be used to prune all the G - hypotheses that do not cover the positive example. </li></ul>G = {h1, h21, …, hn} S = { h1’, h3’, …,hm’ } Don’t cover the last positive example
  38. 38. Eliminate redundant hypotheses <ul><li>If a hypothesis from G is more specific than another hypothesis from G : eliminate it ! </li></ul>More specific than another general hypothesis Obviously also for S ! G = {h1, h21, …, hn} S = { h1’, h3’, …,hm’ } Reason: Invariant acts as a wave front : anything above G is not allowed. The most general elements of G define the real boundary
  39. 39. Convergence: <ul><li>Eventually, if G and S MAY get a common element: Version Spaces has converged to a solution . </li></ul><ul><li>Remaining examples need to be verified for the solution . </li></ul>G = {h1, h21, h22, …, hn} S = { h13’, h2’, …,hm’ }
  40. 40. Reaction example <ul><li>Initialization: </li></ul>[ ?, ?, ?, ?] Most general  Most specific
  41. 41. Alma3, breakfast, Friday, cheap: + <ul><li>Positive example: minimal generalization of  </li></ul>[ ?, ?, ?, ?]  [Alma3, breakfast, Friday, cheap]
  42. 42. DeMoete, lunch, Friday, expensive: - <ul><li>Negative example: minimal specialization of [ ?, ?, ?, ?] </li></ul><ul><li>15 possible specializations !! </li></ul>[ Alma3 , ?, ?, ?] [ DeMoete , ?, ?, ?] [ Sedes , ?, ?, ?] [?, breakfast , ?, ?] [?, lunch , ?, ?] [?, dinner , ?, ?] [?, ?, Monday , ?] [?, ?, Tuesday , ?] [?, ?, Wednesday , ?] [?, ?, Thursday , ?] [?, ?, Friday , ?] [?, ?, Saturday , ?] [?, ?, Sunday , ?] [?, ?, ?, cheap ] [?, ?, ?, expensive ] Matches the negative example X X X X Does not generalize the specific model Specific model: [Alma3, breakfast, Friday, cheap] X X X X X X X X Remain !
  43. 43. Result after example 2: [ ?, ?, ?, ?]  [Alma3, breakfast, Friday, cheap] [ Alma3 , ?, ?, ?] [?, breakfast , ?, ?] [?, ?, ?, cheap ]
  44. 44. Alma3, lunch, Saturday, cheap: + <ul><li>Positive example: minimal generalization of [Alma3, breakfast, Friday, cheap] </li></ul>[Alma3, breakfast, Friday, cheap] [Alma3, ?, ?, ?] [?, breakfast, ?, ?] [?, ?, ?, cheap] [Alma3, ? , ? , cheap] does not match new example
  45. 45. Sedes, breakfast, Sunday, cheap: - <ul><li>Negative example: minimal specialization of the general models: </li></ul>The only specialization that is introduced is pruned, because it is more specific than another general hypothesis [Alma3, ?, ?, ?] [?, ?, ?, cheap] [Alma3, ? , ? , cheap] [ Alma3 , ?, ?, cheap]
  46. 46. Alma 3, breakfast, Sunday, expensive: - <ul><li>Negative example: minimal specialization of [Alma3, ?, ?, ?] </li></ul>Cheap food at Alma3 produces the allergy ! [Alma3, ?, ?, ?] [Alma3, ? , ? , cheap] [Alma3, ?, ?, cheap ] Same hypothesis !!!
  47. 47. Version Space Algorithm: Generalize all hypotheses in S that do not cover the example yet, but ensure the following: - Only introduce minimal changes on the hypotheses. - Each new specific hypothesis is a specialization of some general hypothesis . - No new specific hypothesis is a generalization of some other specific hypothesis . Prune away all hypotheses in G that do not cover the example. Initially : G := { the hypothesis that covers everything} S := {  } For each new positive example:
  48. 48. Version Space Algorithm (2): Specialize all hypotheses in G that cover the example, but ensure the following: - Only introduce minimal changes on the hypotheses. - Each new general hypothesis is a generalization of some specific hypothesis . - No new general hypothesis is a specialization of some other general hypothesis . Prune away all hypotheses in S that cover the example. Until there are no more examples: report S and G OR S or G become empty : report failure . . . For each new negative example:
  49. 49. Properties of VS: <ul><li>Symmetry: </li></ul><ul><ul><li>positive and negative examples are dealt with in a completely dual way. </li></ul></ul><ul><li>Does not need to remember previous examples. </li></ul><ul><li>Noise: </li></ul><ul><ul><li>VS cannot deal with noise ! </li></ul></ul><ul><ul><ul><li>If a positive example is given to be negative then VS eliminates the desired hypothesis from the Version Space G ! </li></ul></ul></ul>
  50. 50. Termination: <ul><li>If it terminates because of “no more examples”: </li></ul>Then all these hypotheses , and all intermediate hypotheses , are still correct descriptions covering the test data. <ul><ul><li>Example (spaces on termination): </li></ul></ul>[Alma3, ?, ?, ?] [?, ?, Monday,?] [Alma3, ?, Monday, cheap] [Alma3, ?, ?, cheap] [?, ?, Monday, cheap] [Alma3, ?, Monday, ?] VS makes NO unnecessary choices !
  51. 51. Termination (2): <ul><li>If it terminates because of S or G being empty: </li></ul><ul><ul><li>Then either: </li></ul></ul><ul><ul><ul><li>The data is inconsistent (noise?) </li></ul></ul></ul><ul><ul><ul><li>The target concept cannot be represented in the hypothesis-language H. </li></ul></ul></ul>[Alma3, breakfast, ?, cheap]  [Alma3, lunch, ?, cheap] <Alma3, dinner, Sunday, cheap> - <Alma3, breakfast, Sunday, cheap> + <Alma3, lunch, Sunday, cheap> + <ul><li>Example: </li></ul><ul><ul><li>target concept is: </li></ul></ul><ul><ul><li>given examples like: </li></ul></ul><ul><ul><li>this cannot be learned in our language H. </li></ul></ul>
  52. 52. Which example next? <ul><li>VS can decide itself which example would be most useful next. </li></ul><ul><ul><li>It can ‘query’ a user for the most relevant additional classification ! </li></ul></ul><ul><li>Example: </li></ul><Alma3,lunch,Monday, expensive > [Alma3, ?, ?, ?] [?, ?, Monday,?] [Alma3, ?, Monday, cheap] [Alma3, ?, ?, cheap] [?, ?, Monday, cheap] [Alma3, ?, Monday, ?] classified negative by 3 hypotheses classified positive by 3 hypotheses is the most informative new example
  53. 53. Use of partially learned concepts: <ul><li>Example: <Alma3,lunch,Monday,cheap> </li></ul>[Alma3, ?, ?, ?] [?, ?, Monday,?] [Alma3, ?, Monday, cheap] [Alma3, ?, ?, cheap] [?, ?, Monday, cheap] [Alma3, ?, Monday, ?] <ul><ul><li>can be classified as positive </li></ul></ul><ul><ul><ul><li>is covered by all remaining hypotheses ! </li></ul></ul></ul><ul><ul><ul><li>it is enough to check that it is covered by the hypotheses in S ! (all others generalize these) </li></ul></ul></ul>
  54. 54. Use of partially learned concepts (2): <ul><li>Example: <Sedes,lunch,Sunday,cheap> </li></ul>[Alma3, ?, ?, ?] [?, ?, Monday,?] [Alma3, ?, Monday, cheap] [Alma3, ?, ?, cheap] [?, ?, Monday, cheap] [Alma3, ?, Monday, ?] <ul><ul><li>can be classified as negative </li></ul></ul><ul><ul><ul><li>is not covered by any remaining hypothesis ! </li></ul></ul></ul><ul><ul><ul><li>it is enough to check that it is not covered by any hypothesis in G ! (all others specialize these) </li></ul></ul></ul>
  55. 55. Use of partially learned concepts (3): <ul><li>Example: <Alma3,lunch,Monday,expensive> </li></ul>[Alma3, ?, ?, ?] [?, ?, Monday,?] [Alma3, ?, Monday, cheap] [Alma3, ?, ?, cheap] [?, ?, Monday, cheap] [Alma3, ?, Monday, ?] <ul><ul><li>can not be classified </li></ul></ul><ul><ul><ul><li>is covered by 3, not covered 3 hypotheses </li></ul></ul></ul><ul><ul><ul><ul><ul><li>no conclusion </li></ul></ul></ul></ul></ul>
  56. 56. Use of partially learned concepts (4): <ul><li>Example: <Sedes,lunch,Monday,expensive> </li></ul>[Alma3, ?, ?, ?] [?, ?, Monday,?] [Alma3, ?, Monday, cheap] [Alma3, ?, ?, cheap] [?, ?, Monday, cheap] [Alma3, ?, Monday, ?] probably does not belong to the concept : ratio : 1/6 <ul><ul><li>can only be classified with a certain degree of precision </li></ul></ul><ul><ul><ul><li>is covered by 1, not covered 5 hypotheses </li></ul></ul></ul>
  57. 57. The relevance of inductive BIAS: choosing H <ul><li>Our hypothesis language L fails to learn some concepts. </li></ul><ul><li>What about choosing a more expressive language H? </li></ul><ul><li>Assume H’: </li></ul><ul><ul><li>See example : [Alma3, breakfast, ?, cheap]  [Alma3, lunch, ?, cheap] </li></ul></ul><ul><ul><li>allows conjunctions (as before) </li></ul></ul><ul><ul><li>allows disjunction and negation too ! </li></ul></ul><ul><ul><ul><li>Example: </li></ul></ul></ul><ul><ul><ul><ul><li>Restaurant = Alma3  ~(Day = Monday) </li></ul></ul></ul></ul>
  58. 58. Inductive BIAS (2): <ul><li>This language H’ allows to represent ANY subset of the complete set of all events X </li></ul><ul><li>But X has 126 elements </li></ul><ul><ul><li>we can express 2 126 different hypotheses now ! </li></ul></ul>Restaurant Meal Day Cost 3 X 3 X 7 X 2 = 126
  59. 59. Inductive BIAS (3) <ul><li>Version Spaces using H’: </li></ul>Restaurant Meal Day Cost Reaction Alma 3 breakfast Friday cheap Yes De Moete lunch Friday expensive No Alma 3 lunch Saturday cheap Yes Sedes breakfast Sunday cheap No Alma 3 breakfast Sunday expensive No + - + - - ?  {~[DeMoete. Lunch, Friday, expensive]  ~[Sedes, breakfast, Sunday, cheap]} {~[DeMoete. Lunch, Friday, expensive]  ~[Sedes, breakfast, Sunday, cheap]  ~[Alma 3, breakfast, Sunday, expensive]} {~[DeMoete. Lunch, Friday, expensive]} {[Alma 3, breakfast, Friday, cheap]  [Alma 3, lunch, Saturday, cheap]} {[Alma 3, breakfast, Friday, cheap]}
  60. 60. Inductive BIAS (3) <ul><li>Resulting Version Spaces: </li></ul><ul><li>We haven’t learned anything </li></ul><ul><ul><li>merely restated our positive and negative examples ! </li></ul></ul>G = {~[DeMoete. Lunch, Friday, expensive]  ~[Sedes, breakfast, Sunday, cheap]  ~[Alma 3, breakfast, Sunday, expensive]} S = {[Alma 3, breakfast, Friday, cheap]  [Alma 3, lunch, Saturday, cheap]} <ul><li>In general: in order to be able to learn, we need an inductive BIAS (= assumption): </li></ul><ul><ul><li>Example: “ The desired concept CAN be described as a conjunction of features “ </li></ul></ul>
  61. 61. Shift of Bias <ul><li>Practical approach to the Bias problem: </li></ul><ul><li>Avoids the choice. </li></ul><ul><li>Gives the most general concept that can be learned. </li></ul><ul><ul><ul><li>Start VS with a very weak hypothesis language. </li></ul></ul></ul><ul><ul><ul><li>If the concept is learned: o.k. </li></ul></ul></ul><ul><ul><ul><li>Else </li></ul></ul></ul><ul><ul><ul><ul><li>Refine your language and restart VS </li></ul></ul></ul></ul>

×