ProbabilisticCircuits.jl
ProbabilisticCircuits.jl copied to clipboard
Tractables
Probabilistic Circuits from the Juice library
1. Given Logical Formula, and a psdd (pc, vtree) 2. Compile it to a smooth logical circuit with the same vtree 3. Compute Expectation of logical circuit w.r.t. the pc....
Having multiple distribution type for inputs fails gpu kernels due to Union split limiation. Currently, can have at most 3 types in the Union of a CuArray for cuda kernels....
- [ ] Param Learning - [x] EM - [ ] EM CPU - [ ] Regurlarization (Softness) - [ ] SGD version. Best option seems to be to customize...
I have a situation where I need to train models with lots of missing information, i.e. the training matrix is very sparse (on the order of more than 90% of...
Most of our inference functions assume something about the circuit (smoothness, decomposability, determinism, structured decomposability). Currently there are some functions that check these preconditions explicitly in code, some that add...
Hi, Here's a MWE (up-to-date with both `master` branches of PCs and LCs): ```julia # Utility function to create literals f(x) = PlainProbLiteralNode.(Int32.([-x, x])) # Bernoulli parameters W = map(x...
We have some randomness in our unit tests, would be useful for reproducing errors that come up if we could have a setup so that we log/print the seed somehow...
Benchmark common queries on few different circuit sizes, batch sizes, and cpu vs gpu, etc. - Circuit Size [track nodes, edges, layers in bitcircuit] - Batch size (when releveant) [1,...
1. Easy CON API Currently user has to manually call MAR twice to get conditionals. Have some api like this: ``` conditional(circuit::ProbCircuit, data::DataFrame, q_indices) ``` Where `q_indice`s give indexes of...
Cf https://arxiv.org/pdf/1009.2722.pdf Cf https://libra.cs.uoregon.edu/ ?