locked
inferred probabilities RRS feed

  • Question

  • Hi,

    defined a model which I infer a binary discrete variable (range = 2).   in some cases it happened that the sum of the probabilities of the two states was not 1.I am attaching two Example, one  correctly inferred values ​​and those in the wrong way.

    prob_p0	prob_p1
    5.01E-66	6.23E-67
    1.53E-67	1.16E-70
    9.12E-49	1.42E-43
    4.25E-47	1.10E-48
    3.14E-78	1.22E-75
    1.59E-46	5.98E-50
    3.25E-39	2.87E-48
    4.19E-56	1.27E-52
    2.97E-43	3.83E-34
    1.33E-51	2.59E-49
    1.02E-35	1.25E-42
    7.41E-46	2.63E-43
    1.07E-42	4.14E-32
    8.00E-63	6.55E-70
    5.24E-35	2.24E-40
    9.50E-46	3.69E-41
    2.41E-64	7.36E-66
    1.15E-46	5.41E-55
    2.05E-55	2.62E-47
    1.62E-54	3.62E-55
    2.21E-42	3.40E-45
    
    Correct
    prob_p0	prob_p1
    1.36E-82	1
    5.95E-78	1
    2.30E-75	1
    4.14E-38	1
    2.47E-77	1
    2.89E-74	1
    1.53E-60	1
    4.43E-53	1
    5.94E-70	1
    1.24E-80	1
    1.15E-79	1
    3.45E-49	1
    1.16E-73	1
    1.45E-35	1
    1.26E-48	1
    7.55E-49	1
    6.73E-60	1
    3.07E-56	1
    2.79E-66	1
    2.23E-39	1
    4.95E-82	1
    


    Tuesday, July 19, 2011 8:33 AM

All replies

  • Hi Alessandro

    What is your model and data?

    Tuesday, July 19, 2011 2:35 PM
    Owner
  • I implemented three models:
    - Naive Bayes
    - Tree augmented naive Bayes (TAN)
    - Hidden Naive Bayes

    in all three models I have this problem with different frequency

    the dataset is composed of a thousand files of different lengths.

    As soon as I find a stable error I will share the code.
    Wednesday, July 20, 2011 11:32 AM