Coverage for bmm_multitask_learning/coalescent/parameters_cov.py: 0%

1import numpy as np 

2 

3from .coalescent_inference import CoalescentTree 

4 

5 

6def optimal_cov(coalesce_tree: CoalescentTree, dim): 

7 """ 

8 retunrns samples from covariance matrix 

9 for InverseWishart distribution parameters update 

10 """ 

11 sample_mean = np.zeros((dim, ), dtype=float) 

12 overall_items = 0 

13 

14 # compute sample mean 

15 for item in coalesce_tree.iterate_over(): 

16 parent = item.parent 

17 if parent is not None: 

18 overall_items += 1 

19 dt = item.t - parent.t 

20 D_i = item.mean - parent.mean 

21 x_i = D_i / np.sqrt(dt) 

22 sample_mean += x_i 

23 sample_mean /= overall_items 

24 

25 

26 # compute samples 

27 samples = [] 

28 

29 for item in coalesce_tree.iterate_over(): 

30 parent = item.parent 

31 if parent is not None: 

32 dt = item.t - parent.t 

33 D_i = item.mean - parent.mean 

34 x_i = D_i / np.sqrt(dt) 

35 samples.append(x_i - sample_mean) 

36 

37 return np.array(samples) 

38 

39 

40def optimal_R(s_leaves, weights): 

41 y = np.exp(- s_leaves) * weights # n x d 

42 return y