gaussian

Prototyping Samplers for GTSAM

Now largely built into gtsam.discrete.

---

source

sample_bayes_net

def sample_bayes_net(
    bn:GaussianBayesNet, N:int
)->dict:

High performance ancestral sampling. It returns a dictionary of nj x N samples, where n_j is the dimensionality for key j.

---

source

sample_conditional

def sample_conditional(
    node:GaussianConditional, N:int, parents:list=[], sample:dict={}
):

Sample from conditional

_x_, _y_ = 11, 12
bayesNet = gtsam.GaussianBayesNet()
I_1x1 = np.eye(1, dtype=float)
bayesNet.push_back(gtsam.GaussianConditional(_x_, [9.0], I_1x1, _y_, I_1x1))
bayesNet.push_back(gtsam.GaussianConditional(_y_, [5.0], I_1x1))

sample_bayes_net(bayesNet, 4)

{12: array([[2.83568484, 5.93936955, 3.94485894, 6.90187555]]),
 11: array([[5.2819049 , 2.60654641, 4.86926211, 0.530073  ]])}