Result of Applying Noisy Harmonic Grammar to 1SpreadsheetForTableau1.xlsx

12-15-2021, 5:22 p.m.

OTSoft 2.6, release date 4/11/2021

1. Weights Found

4.954        Prefer Majority
4.000        Prefer Minority

2. Matchup to Input Frequencies


   /Input1/   Input Fr. Gen Fr.  Input #     Gen. #
   Majority     0.750   0.749                749363
   Minority     0.250   0.251                250637

/Input1/: 
                  Prefer Majority (4.954)|Prefer Minority (4)
>Majority (0.749)                        |        1          
 Minority (0.251)           1            |                   


3. Testing the Grammar:  Details

The grammar was tested for 1000000 cycles.

Average error per candidate:  0.000 percent

Log likelihood of the data: -0.562

Learning time:  0.000 minutes



4. Parameter Values Used


   The algorithm used was the Classical version of Noisy Harmonic Grammar (Boersma and Pater 2013).

   Noise was set to have a standard deviation of 1.

   Negative weights were not permitted.
   The number of times that a negative weight was clipped during learning and testing was 41 out of 2000000 chances.

   The number of ties between winner and rival was 0 out of 2000000 chances.
   Ties were resolved by dropping the trial in question.

   Noise was applied to constraints, hence the same for all candidates.

   Noise was applied before weights multiplied by violations.

   Initial Weights (final weights shown in parentheses)

4.954        4.954   4.954 (4.954)   Prefer Majority
4.000        4.000   4.000 (4.000)   Prefer Minority

   Schedule for Plasticity

      Stage   Trials   PlastMark  PlastFaith 
     1        0          0.001      0.001      
     2        0          0.000      0.000      
     3        0          0.000      0.000      
     4        0          0.000      0.000      

      There were a total of 0 learning trials.
