Binomial Input Nodes

[khosravipasha]

Intitial Code for binomial input nodes, learning and sampling (on gpu) now works (see examples/binomial_mnist.jl).

Todos:

• MAP GPU (debug; was throwing an error in the bigger circui, the tests are passing though)
• How to handle missing flow? I think I might have wrong formula.
• Add somes Tests
• Save/Load for jpc format
• Sample on CPU
• MAP CPU

[codecov]

Codecov Report

Merging #119 (c50f1a2) into master (0e28de4) will increase coverage by 1.18%.
The diff coverage is 53.21%.

@@            Coverage Diff             @@
##           master     #119      +/-   ##
==========================================
+ Coverage   41.94%   43.12%   +1.18%     
==========================================
  Files          22       23       +1     
  Lines        1726     1834     +108     
==========================================
+ Hits          724      791      +67     
- Misses       1002     1043      +41     

Impacted Files	Coverage Δ
src/bits_circuit.jl	87.30% <0.00%> (-0.40%)	⬇️
src/io/plot.jl	88.88% <0.00%> (-11.12%)	⬇️
src/nodes/categorical_dist.jl	23.15% <0.00%> (+0.81%)	⬆️
src/nodes/input_distributions.jl	6.66% <0.00%> (ø)
src/queries/likelihood_cpu.jl	59.61% <0.00%> (ø)
src/structures/hclts.jl	74.72% <0.00%> (-1.42%)	⬇️
src/nodes/binomial_dist.jl	54.87% <54.87%> (ø)
src/io/jpc_io.jl	99.12% <100.00%> (+0.11%)	⬆️
src/nodes/abstract_nodes.jl	88.00% <0.00%> (+0.24%)	⬆️
... and 6 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0e28de4...c50f1a2. Read the comment docs.