Week 1 Assignment - Section 5

Course Q&A Math for Machine Learning Probability & Statistics for Machine Learning &...
1

I’m getting the following results:
Distribution parameters for training split:

{0: {‘bark_days’: params_binomial(n=30.000, p=0.800),
‘ear_head_ratio’: params_uniform(a=0.100, b=0.597),
‘height’: params_gaussian(mu=34.955, sigma=1.477),
‘weight’: params_gaussian(mu=19.916, sigma=0.997)},
1: {‘bark_days’: params_binomial(n=30.000, p=0.497),
‘ear_head_ratio’: params_uniform(a=0.201, b=0.500),
‘height’: params_gaussian(mu=30.041, sigma=1.999),
‘weight’: params_gaussian(mu=24.910, sigma=5.020)},
2: {‘bark_days’: params_binomial(n=30.000, p=0.298),
‘ear_head_ratio’: params_uniform(a=0.101, b=0.300),
‘height’: params_gaussian(mu=40.152, sigma=3.520),
‘weight’: params_gaussian(mu=31.927, sigma=3.045)}}

Probability of each class for training split:

{0: 0.346, 1: 0.393, 2: 0.26}

I should be getting:

Expected Output 
Distribution parameters for training split:

{0: {'bark_days': params_binomial(n=30.000, p=0.801),
     'ear_head_ratio': params_uniform(a=0.100, b=0.597),
     'height': params_gaussian(mu=35.030, sigma=1.518),
     'weight': params_gaussian(mu=20.020, sigma=1.012)},
 1: {'bark_days': params_binomial(n=30.000, p=0.498),
     'ear_head_ratio': params_uniform(a=0.201, b=0.500),
     'height': params_gaussian(mu=29.971, sigma=2.010),
     'weight': params_gaussian(mu=24.927, sigma=5.025)},
 2: {'bark_days': params_binomial(n=30.000, p=0.296),
     'ear_head_ratio': params_uniform(a=0.101, b=0.300),
     'height': params_gaussian(mu=39.814, sigma=3.572),
     'weight': params_gaussian(mu=31.841, sigma=3.061)}}

Probability of each class for training split:

{0: 0.346, 1: 0.393, 2: 0.26}

I think this is the cause for why my final accuracy score is only 0.98 instead of 1.00. Is this just a rounding thing that can be ignored or something more significant?

2

Hi @Ralph_Nussbaumer, it looks to me that your generated data might be slightly different from what the assignment expects. If you want, send me your notebook through a DM and I will take a look.

3

Hi a-zarta, I figured it out. I was not calling the uniform_generator function from the generator functions further down. I simply used the same code as in the generator functions as I used in uniform_generator. The bolded line from the instructions to be sure use functions defined earlier gave me the hint. I guess the np.random.seed(42) must initialize the data somehow to make it sync with what is expect. Thanks for the offer to help

Ralph

4

Cool! Indeed that is the reason why we ask to use uniform_generator function.