Feras Saad

It could be useful to have a Python command-line interface for SPPL, i.e., ```bash $ pyhton -m sppl compile # produces model.spe $ python -m sppl logprob $ python -m...

The minimal example is a 2-mixture of linear regression, with known coefficients and a single real-covariate ``` func(x): U ~ bernoulli(.3) if U == 0: X ~ normal(2 * x...

Implement multi-variable environment in spn.Product

1

https://github.com/probcomp/sum-product-dsl/blob/c482341d04ef35febefc46f8821350e9236e8013/src/spn.py#L409-L413 1. The multivariable transform must be of type event. 2. If a condition or logprob query has multivariate symbol, then we cannot call logprob_factored on the children. Instead, we...

Fix Reciprocal.ffwd in case that argument is not an integer (sympy chokes)

2

https://github.com/probcomp/sum-product-dsl/blob/95c0daf08af9eea974e0053a45ec9d462e18dbfc/src/transforms.py#L593 Test case ``` >>> spn = X >> norm(0,1) >>> spn.transform(Z, 2**(1/X)) >>> samples = spn.sample(10) ``` The final call will cause sympy to hang, which cannot evaluate two...

Example, the following variable cannot be rewritten in sample_func ```python X = Variable('X[0]') sample_func(lambda ??: ??, 100) ``` One solution is to handle array variables by naming them `X_at_10` in...