This is an archived version of the Machine Learning Lab website for documentation purposes. The host is in no way affiliated with the University of Freiburg, its Faculty of Engineering or any of its members.

Machine Learning Exercises Summer Term 2010

Contact

For questions and general comments please contact the exercise course organizer Dr. Thomas Gabel.

Top Downloads

  • Slides of the award ceremony (including the final ranking) of the course-internal data-mining competition: slides
  • The perceptron learning program used for solving exercise 6: link.
  • The stand-alone presentation on learning an aggressive defense behavior (exercise 15).
  • The solution presentations of all participants of our data-mining competition: link
  • CSV data files for exercise 4: disease data and body height data

All downloads (including sample solutions for all exercises) can be found here.

Course-Internal Data-Mining Competition

Task Details: see Exercise Sheet 2

News

Due to numerous requests, the submission deadline has been extended to June, 27th, 2010.

General notes and reminders:

  • Don't forget to submit your solution (by email to Thomas Gabel) until June, 20th, 2010 June, 27th, 2010.
  • That email must also list the names of all your team members and corresponding matriculation numbers.
  • You can work in teams with up to 3 students.
  • Your solution must be submitted in the form of a list with entries as follows:
    <Customer number>;<Voucher>
  • The voucher field must contain a “0” for “no voucher” or a “1” for “send voucher”. The “Customer number” field must contain the entries given in file dmc2010_class.txt.
  • Results that do not contain all the customer numbers from the data that is to be assessed will not be accepted.

Furthermore, you are required to present your solution in one of the the exercise sessions following the submission deadline.

  • Presentation length: 15 minutes (equally distributed among team members!).
  • Contents of your talk:
    • explain briefly the foundations of the methods you employed and tell the other why you decided to use those methods
    • speak about the steps you performed solving the task, including possible missteps or problems encountered
    • present the results you obtained, also using appropriate charts and diagrams.
  • The schedule of the presentations will be listed on this web page after the submission deadline.
    • Hand in the slides of your presentation prior to your talk (by email to Thomas Gabel).
  • Check this page frequently for any updates or changes.

Assessment of your solution: Your solution will be considered successful only, if you comply to the following five minimal requirements.

  • You have submitted your solution before the submission deadline.
  • You have covered all customer numbers given in dmc2010_class.txt (not just a subset) and your solution is formatted as described above.
  • Your solution is not a trivial one (like, for example, sending vouchers to all customers).
  • Your presentation is convincing in that it (a) shows that you have seriously dealt with the task and (b) the things you are presenting are in alignment with your submitted solution.
  • You have handed in the slides of your presentation prior to your talk.
  • Besides those minimal requirements (which are essential for exam qualification), we will have a course-internal DMC ranking.
    • In order to determine this ranking, the cost structure given in the task specification will be employed.
    • The best three teams will be awarded.

Schedule of Result Presentations

UE5 (13.07.2010):

  1. Thomas Huber and Nichola Abdo and Mesbah Anwari
  2. Sebastian Ohse
  3. Stefan Osswald
  4. Silvan Sievers and Jens Silva and Matthias Frorath
  5. Philipp Blohm and Roxana Bersan
  6. Manuel Braun and Jendrik Seipp

UE6 (20.07.2010):

  1. Thorsten Engesser and Jeremias Holub
  2. Benedikt Mendorf
  3. Edgar Oswald
  4. Simon Pahl and Jan Kelch and Meike Aulbach

UE7 (22.07.2010):

  1. Hernando Ramirez and Dragos-Alexandru Sorescu and Suzana Ilinca Tudose
  2. Kerstin Schmatzer and Rufino Khoun and Thorsten Bischler
  3. Johannes Stegmaier and Nisha Rajasundaram
  4. Yousra Wahby and Alejandro Molina Ramirez and Refik Hadzialic
  5. Richard Zahoransky and Alexander Schneider
  6. Friedel Ziegelmayer and Daniel Christiany and Marcus Zeiger
  7. Thomas Gabel:
    Award Ceremony: Proclamation of the official course-internal DMC ranking and presentation of prizes.

Final Ranking of the Course-Internal Data-Mining Competition

The baseline for the test data set (sending a voucher to all customers) is 8548.5. Therefore, the top seven teams yield a monetary gain; with the winning team earning almost 1400 Euro. Congratulations!

  1. bersan_blohm, 9931.0
  2. pahl_kelch_aulbach, 9366.5
  3. ramirez_tudose_sorescu, 9012.5
  4. braun_seipp, 8927.5
  5. stegmaier_rajasundaram, 8726.0
  6. schmatzer_bischler_khoun, 8591.0
  7. sievers_silva_frorath, 8567.0
  8. wahby_ramirez_hadzialic, 8369.5
  9. ziegelmayer_christiany_zeiger, 7688.0
  10. oswald, 7573.5
  11. ohse, 4419.0
  12. mendorf, 2116.0
  13. osswald, 762.5
  14. zahoransky_schneider, 319.0
  15. engesser_holub, -29.5
  16. huber_abdo_anwari, NaN

Additional Solutions (handed in after the deadline, e.g. containing error corrections etc.). Interestingly, the best solution handed in outperforms the baseline by more than 2200 Euro.

  • braun_seipp (2), 10768.0
  • braun_seipp (1), 10759.0
  • engesser_holub, 9746.5
  • bersan_blohm, 9426.5
  • huber_abdo_anwari, 8766.5

You will find more information on the results in this set of slides.

Course-Intenal Ranking Blended with the Official DMC Ranking

  1. Inst_KIT_2 12545,5
  2. TU_Dortmund_2.txt 12445,5
  3. Inst_KIT_1 12377,5
  4. RWTH_Aachen_1 12119,5
  5. FH_Trier_1 12072,5
  6. RWTH_Aachen_2 12072,0
  7. Uni_Marburg_1 12006,5
  8. HS_Anhalt_2 11961,5
  9. HS_Anhalt_1 11941,5
  10. Uni_UPC Barcelona_1 11566,0
  11. Uni_Hamburg_2 11532,5
  12. FH_Köln_1 11444,0
  13. HS_Wismar_1 11423,0
  14. Uni_Hamburg_1 11409,5
  15. FH_Lausitz_1 11349,5
  16. Uni_Bloominton_1.txt 11285,0
  17. HS_Mittweida_1 11275,5
  18. Uni_Science Technology China_2 11253,5
  19. HS_Wismar_2 11169,5
  20. TU_Berlin_1 11049,0
  21. Uni_Ljubljana_1 10834,0
    ADDITIONAL_braun_seipp (2), 10768.0
  22. Uni_Brigham_1 10759,0
    ADDITIONAL_braun_seipp (1), 10759.0
  23. HS_Ulm_1 10569,0
  24. X_HS_Mittweida_2 10489,5
  25. Uni_Zagreb_1 10461,5
  26. Uni_Kiel_1 10145,0
  27. Uni_Leibniz Hannover_1 10034,0
    bersan_blohm, 9931.0
  28. U_Potsdam_1 9883,5
  29. FH_Gelsenkirchen_1 9827,5
    ADDITIONAL_engesser_holub, 9746.5
  30. Uni_Gadjah Mada_2 9503,5
    ADDITIONAL_bersan_blohm, 9426.5
    pahl_kelch_aulbach, 9366.5
    ramirez_tudose_sorescu, 9012.5
    braun_seipp, 8927.5
  31. FH_Schmalkalden_2 8838,0
    ADDITIONAL_huber_abdo_anwari, 8766.5
  32. Uni_Ljubljana_2 8738,5
    stegmaier_rajasundaram, 8726.0
  33. TU_Dortmund_1 8720,5
  34. Uni_Gadjah Mada_1 8668,0
    schmatzer_bischler_khoun, 8591.0
    sievers_silva_frorath, 8567.0
    BASELINE, 8548.5
    wahby_ramirez_hadzialic, 8369.5
  35. FH_Hannover_1 8262,0
  36. HS_Reutlingen_1 7799,0
  37. FH_Lausitz_2 7713,0
    ziegelmayer_christiany_zeiger, 7688.0
  38. Uni_Zürich_1 7630,0
  39. Uni_Dublin_1 7614,5
    oswald, 7573.5
  40. Inst_Telkom_1 7516,5
  41. HS_Wirtschaft Recht Berlin_1.txt 7401,5
  42. HS_Reutlingen_2 7295,5
  43. FH_Koblenz_1 5934,5
  44. HS_Ulm_2 4647,0
    ohse, 4419.0
  45. Inst_Teknologi Sepuluh_1 4078,0
  46. Uni_Athen Economics Business_1 3008,5
    mendorf, 2116.0
  47. Fernuni_Hagen_2.txt 1115,0
  48. Uni_Isfahan_1 865,5
    osswald, 762.5
    zahoransky_schneider, 319.0
  49. Inst_Telkom_2 0
    engesser_holub, -29.5
  50. Uni_Valparaiso_1 -149,0
  51. Uni_Valparaiso_2 -173,0
  52. Uni_Lviv Polytechnic_1 -192,5
  53. FH_Muenster_1 -260,5
  54. FH_Erfurt_1.txt -710,0
  55. Uni_Laval_1 -2191,5
  56. Uni_Thessaloniki_1 -2634,0
  57. Uni_Thessaloniki_2 -2889,0
  58. Inst_Beijing_1 -3452,0
  59. FH_Hannover_2 NaN
    huber_abdo_anwari, NaN
  60. TU_Dresden_1 NaN
  61. Uni_Bologna_1 NaN
  62. Uni_Eindhoven_1 NaN
  63. Uni_Eindhoven_2 NaN
  64. Uni_Perm_1 NaN
  65. Uni_RFW Bonn_1 NaN
  66. Uni_Wroclaw_1 NaN