notebook.community

Edit and run 





In [1]:



    


from scipy.special import expit
from rbm import RBM
from sampler import VanillaSampler, PartitionedSampler
from trainer import VanillaTrainier
from performance import Result
import numpy as np
import datasets, performance, plotter, mnist, pickle, rbm, os, logging

logger = logging.getLogger()
# Set the logging level to logging.DEBUG to 
logger.setLevel(logging.INFO)


%matplotlib inline

models_names = [
    "one","two","three","four","five","six","seven", "eight", "nine", "bar","two_three"]
# RBM's keyed by a label of what they were trained on
models = datasets.load_models(models_names)



    











In [18]:



    


data_set_size = 40
number_gibbs_alternations = 1000
# the model we will be `corrupting` the others with, in this case we are adding bars to the digit models
corruption_model_name = "bar"

def result_key(data_set_size, num_gibbs_alternations, model_name, corruption_name):
    return '{}Size_{}nGibbs_{}Model_{}Corruption'.format(data_set_size, number_gibbs_alternations, model_name, corruption_name)

def results_for_models(models, corruption_model_name, data_set_size, num_gibbs_alternations):
   
    results = {}
    for model_name in models:
        if model_name is not corruption_model_name:
            key = result_key(data_set_size, number_gibbs_alternations, model_name, corruption_model_name)
            logging.info("Getting result for {}".format(model_name))
            model_a = models[model_name]   
            model_b = models[corruption_model_name]
            model_a_data = model_a.visible[:data_set_size]#visibles that model_a was fit to. 
            model_b_data = model_b.visible[:data_set_size]#visibles that model_b was fit to. 
            r = Result(data_set_size, num_gibbs_alternations, model_a, model_b,model_a_data, model_b_data)
            r.calculate_result()
            results[key] = r
    return results
    
results = results_for_models(models, corruption_model_name, data_set_size, number_gibbs_alternations)



    


















    






INFO:root:Getting result for nine










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for seven










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for four










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for two










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for eight










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for five










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for one










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






/Users/Max/Documents/Uni/400-Level/ENGR489/RBM/Max/performance.py:13: RuntimeWarning: divide by zero encountered in log
  score = (target * np.log(sample)) + ((1 - target) * np.log((1 - sample)))
/Users/Max/Documents/Uni/400-Level/ENGR489/RBM/Max/performance.py:13: RuntimeWarning: invalid value encountered in multiply
  score = (target * np.log(sample)) + ((1 - target) * np.log((1 - sample)))
INFO:root:Getting result for three










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for six










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100









    






INFO:root:Getting result for two_three










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 100
Running batch 1 of 100
Running batch 2 of 100
Running batch 3 of 100
Running batch 4 of 100
Running batch 5 of 100
Running batch 6 of 100
Running batch 7 of 100
Running batch 8 of 100
Running batch 9 of 100
Running batch 10 of 100
Running batch 11 of 100
Running batch 12 of 100
Running batch 13 of 100
Running batch 14 of 100
Running batch 15 of 100
Running batch 16 of 100
Running batch 17 of 100
Running batch 18 of 100
Running batch 19 of 100
Running batch 20 of 100
Running batch 21 of 100
Running batch 22 of 100
Running batch 23 of 100
Running batch 24 of 100
Running batch 25 of 100
Running batch 26 of 100
Running batch 27 of 100
Running batch 28 of 100
Running batch 29 of 100
Running batch 30 of 100
Running batch 31 of 100
Running batch 32 of 100
Running batch 33 of 100
Running batch 34 of 100
Running batch 35 of 100
Running batch 36 of 100
Running batch 37 of 100
Running batch 38 of 100
Running batch 39 of 100
Running batch 40 of 100
Running batch 41 of 100
Running batch 42 of 100
Running batch 43 of 100
Running batch 44 of 100
Running batch 45 of 100
Running batch 46 of 100
Running batch 47 of 100
Running batch 48 of 100
Running batch 49 of 100
Running batch 50 of 100
Running batch 51 of 100
Running batch 52 of 100
Running batch 53 of 100
Running batch 54 of 100
Running batch 55 of 100
Running batch 56 of 100
Running batch 57 of 100
Running batch 58 of 100
Running batch 59 of 100
Running batch 60 of 100
Running batch 61 of 100
Running batch 62 of 100
Running batch 63 of 100
Running batch 64 of 100
Running batch 65 of 100
Running batch 66 of 100
Running batch 67 of 100
Running batch 68 of 100
Running batch 69 of 100
Running batch 70 of 100
Running batch 71 of 100
Running batch 72 of 100
Running batch 73 of 100
Running batch 74 of 100
Running batch 75 of 100
Running batch 76 of 100
Running batch 77 of 100
Running batch 78 of 100
Running batch 79 of 100
Running batch 80 of 100
Running batch 81 of 100
Running batch 82 of 100
Running batch 83 of 100
Running batch 84 of 100
Running batch 85 of 100
Running batch 86 of 100
Running batch 87 of 100
Running batch 88 of 100
Running batch 89 of 100
Running batch 90 of 100
Running batch 91 of 100
Running batch 92 of 100
Running batch 93 of 100
Running batch 94 of 100
Running batch 95 of 100
Running batch 96 of 100
Running batch 97 of 100
Running batch 98 of 100
Running batch 99 of 100

















In [24]:



    


for key in models:
#     plotter.plot(results[key].composite)
#     plotter.plot(results[key].visibles_for_stored_hidden(9)[0])
#     plotter.plot(results[key].vis_van_a)
    plotter.plot(models[key].visible[:40])

In the cell below

I have calculated in the previous cell the loglikelyhood score of the partitioned sampling and vanilla sampling technique image-wise. So I have a score for each image. I have done this for all the MNIST digits that have been 'corrupted' by the bar images. That is RBM's trained models 1 - 9 and an RBM trained on 2's and 3's

The wins for a given model are where the partitioned scored better than the vanilla sampling technique

Conversly, losses are images where the vanilla score better.

Intuitively, ties is where they scored the same, which could only really occur when the correction would be zero, or ultimately cancelled out.





In [3]:



    


for key in results:
    logging.info("Plotting, win, lose and tie images for the {}".format(key))
    results[key].plot_various_images()



    


















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_fiveModel_barCorruption










    






Wins For Model A:115 plotting will only show 54 at maximum










    

















    






Losses For Model A: 62 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 223 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_fourModel_barCorruption










    






Wins For Model A:196 plotting will only show 54 at maximum










    

















    






Losses For Model A: 94 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 110 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_sixModel_barCorruption










    






Wins For Model A:190 plotting will only show 54 at maximum










    

















    






Losses For Model A: 29 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 181 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_sevenModel_barCorruption










    






Wins For Model A:204 plotting will only show 54 at maximum










    

















    






Losses For Model A: 106 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 90 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_oneModel_barCorruption










    






Wins For Model A:246 plotting will only show 54 at maximum










    






/Users/Max/Documents/Uni/400-Level/ENGR489/RBM/performance.py:120: RuntimeWarning: invalid value encountered in greater
  win_a = np.compress((part_a > van_a), self.composite, axis = 0)
/Users/Max/Documents/Uni/400-Level/ENGR489/RBM/performance.py:141: RuntimeWarning: invalid value encountered in less
  win_a = np.compress((part_a < van_a), self.composite, axis = 0)










    

















    






Losses For Model A: 13 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 19 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_two_threeModel_barCorruption










    






Wins For Model A:154 plotting will only show 54 at maximum










    

















    






Losses For Model A: 61 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 185 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_threeModel_barCorruption










    






Wins For Model A:296 plotting will only show 54 at maximum










    

















    






Losses For Model A: 66 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 38 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_eightModel_barCorruption










    






Wins For Model A:67 plotting will only show 54 at maximum










    

















    






Losses For Model A: 17 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 316 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_twoModel_barCorruption










    






Wins For Model A:166 plotting will only show 54 at maximum










    

















    






Losses For Model A: 92 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 142 plotting will only show 54 at maximum 










    

















    






INFO:root:Plotting, win, lose and tie images for the 400Size_100nGibbs_nineModel_barCorruption










    






Wins For Model A:183 plotting will only show 54 at maximum










    

















    






Losses For Model A: 42 plotting will only show 54 at maximum 










    

















    






Tie For Model A: 175 plotting will only show 54 at maximum

Thoughts

So on a dataset of size 50, with 100 gibbs alterations we see in all cases that for the digit model, 1,2,3,..,9 that the partitioned sampling technique does either better or the same more often than the vanilla does. Let's try some different configurations.





In [4]:



    


results.update(results_for_models(models, corruption_model_name, 400, 500))



    


















    






INFO:root:Getting result for four










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for seven










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for three










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for two_three










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for nine










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for six










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for two










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for eight










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






INFO:root:Getting result for one










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50









    






/Users/Max/Documents/Uni/400-Level/ENGR489/RBM/performance.py:13: RuntimeWarning: divide by zero encountered in log
  score = (target * np.log(sample)) + ((1 - target) * np.log((1 - sample)))
/Users/Max/Documents/Uni/400-Level/ENGR489/RBM/performance.py:13: RuntimeWarning: invalid value encountered in multiply
  score = (target * np.log(sample)) + ((1 - target) * np.log((1 - sample)))
INFO:root:Getting result for five










    






Constructing Composite Dataset
Generating Vanilla Samples
Generating Partitioned Reconstructions (This may take a while)
Running batch 0 of 50
Running batch 1 of 50
Running batch 2 of 50
Running batch 3 of 50
Running batch 4 of 50
Running batch 5 of 50
Running batch 6 of 50
Running batch 7 of 50
Running batch 8 of 50
Running batch 9 of 50
Running batch 10 of 50
Running batch 11 of 50
Running batch 12 of 50
Running batch 13 of 50
Running batch 14 of 50
Running batch 15 of 50
Running batch 16 of 50
Running batch 17 of 50
Running batch 18 of 50
Running batch 19 of 50
Running batch 20 of 50
Running batch 21 of 50
Running batch 22 of 50
Running batch 23 of 50
Running batch 24 of 50
Running batch 25 of 50
Running batch 26 of 50
Running batch 27 of 50
Running batch 28 of 50
Running batch 29 of 50
Running batch 30 of 50
Running batch 31 of 50
Running batch 32 of 50
Running batch 33 of 50
Running batch 34 of 50
Running batch 35 of 50
Running batch 36 of 50
Running batch 37 of 50
Running batch 38 of 50
Running batch 39 of 50
Running batch 40 of 50
Running batch 41 of 50
Running batch 42 of 50
Running batch 43 of 50
Running batch 44 of 50
Running batch 45 of 50
Running batch 46 of 50
Running batch 47 of 50
Running batch 48 of 50
Running batch 49 of 50

results.update(results_for_models(models, corruption_model_name, 10, 1))





In [6]:



    


results



    


















    
Out[6]:








{'10Size_100nGibbs_eightModel_barCorruption': <performance.Result at 0x10fd90940>,
 '10Size_100nGibbs_fiveModel_barCorruption': <performance.Result at 0x1161b90b8>,
 '10Size_100nGibbs_fourModel_barCorruption': <performance.Result at 0x10fd907b8>,
 '10Size_100nGibbs_nineModel_barCorruption': <performance.Result at 0x10fd90780>,
 '10Size_100nGibbs_oneModel_barCorruption': <performance.Result at 0x10fd90470>,
 '10Size_100nGibbs_sevenModel_barCorruption': <performance.Result at 0x10fd909b0>,
 '10Size_100nGibbs_sixModel_barCorruption': <performance.Result at 0x10fd90be0>,
 '10Size_100nGibbs_threeModel_barCorruption': <performance.Result at 0x10fd90b00>,
 '10Size_100nGibbs_twoModel_barCorruption': <performance.Result at 0x10fd905c0>,
 '10Size_100nGibbs_two_threeModel_barCorruption': <performance.Result at 0x10fd906a0>,
 '400Size_100nGibbs_eightModel_barCorruption': <performance.Result at 0x115b29a20>,
 '400Size_100nGibbs_fiveModel_barCorruption': <performance.Result at 0x115b191d0>,
 '400Size_100nGibbs_fourModel_barCorruption': <performance.Result at 0x1123660f0>,
 '400Size_100nGibbs_nineModel_barCorruption': <performance.Result at 0x1158251d0>,
 '400Size_100nGibbs_oneModel_barCorruption': <performance.Result at 0x115b29128>,
 '400Size_100nGibbs_sevenModel_barCorruption': <performance.Result at 0x115822eb8>,
 '400Size_100nGibbs_sixModel_barCorruption': <performance.Result at 0x115b2b080>,
 '400Size_100nGibbs_threeModel_barCorruption': <performance.Result at 0x114d88080>,
 '400Size_100nGibbs_twoModel_barCorruption': <performance.Result at 0x115b2e9e8>,
 '400Size_100nGibbs_two_threeModel_barCorruption': <performance.Result at 0x115848d68>}

















In [2]:



    


# with open('results_dict', 'wb') as f3le:
#     pickle.dump(results,f3le, protocol = None)
with open('results_dict', 'rb') as f4le:
    results = pickle.load(f4le)



    


















    






---------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
<ipython-input-2-17b5063f203f> in <module>()
      1 # with open('results_dict', 'wb') as f3le:
      2 #     pickle.dump(results,f3le, protocol = None)
----> 3 with open('results_dict', 'rb') as f4le:
      4     results = pickle.load(f4le)

FileNotFoundError: [Errno 2] No such file or directory: 'results_dict'
















In [ ]:



    


# for key in results:
#     if key.startswith('400'):
#         logging.info("Results for hiddens")
#         r = results[key].stored_hiddens
#         for i in range(len(r)):
#             print(results[key].imagewise_score())



    











In [ ]:

Content source: garibaldu/multicauseRBM

Similar notebooks:

notebook.community | gallery | about