v0.4.2

BeliefPropagation v0.4.2

Diff since v0.4.1

Merged pull requests:

• Remove randomness from test (#83) (@stecrotti)
• A few optimizations (#84) (@stecrotti)
• Fix bug with damping (#85) (@stecrotti)
• Test damp on ksat (#86) (@stecrotti)

v0.4.1

BeliefPropagation v0.4.1

Diff since v0.4.0

v0.4.0

BeliefPropagation v0.4.0

Diff since v0.3.0

Breaking changes

Modified the structure of the ProgressAndConvergence callback to store more information

Merged pull requests:

• Improve KSAT (#75) (@stecrotti)
• Store more info about convergence in default Callback (#78) (@stecrotti)
• Add decimation (#79) (@stecrotti)
• Fix possible 0*Inf in maxsum (#80) (@stecrotti)

Closed issues:

• What should we do when messages or beliefs become all zeros? (#77)

v0.3.0

BeliefPropagation v0.3.0

Diff since v0.2.0

Breaking changes

Checking convergence to a fixed point is now a Callback. This changed the signature (only the keyword arguments) of iterate!(bp::BP, ...)

Merged pull requests:

• Reorganize callbacks and convergence checkers (#74) (@stecrotti)

v0.2.0

BeliefPropagation v0.2.0

Diff since v0.1.3

Breaking changes

• The signature of compute_za, compute_zi and compute_zai changed from compute_za(ψₐ, msg_in) to compute_za(bp, a, msg=...) in order to improve the mechanism of dispatching on factor type and message types in optimized versions

Merged pull requests:

• Pass BP object to compute_z* (#70) (@stecrotti)
• Reorganize input arguments for computation of z_* (#71) (@stecrotti)
• Add specific implementation for the KSAT problem (#72) (@stecrotti)

v0.1.3

BeliefPropagation v0.1.3

Diff since v0.1.2

Merged pull requests:

• add dependabot (#61) (@stecrotti)
• Bump julia-actions/cache from 1 to 2 (#62) (@dependabot[bot])
• Bump codecov/codecov-action from 3 to 5 (#63) (@dependabot[bot])
• Bump julia-actions/setup-julia from 1 to 2 (#64) (@dependabot[bot])
• Bump actions/checkout from 3 to 4 (#65) (@dependabot[bot])
• Pass BP object to compute_zai to dispatch not only on message type but model as well (#66) (@stecrotti)
• Add function to initialize messages randomly (#68) (@stecrotti)

v0.1.2

BeliefPropagation v0.1.2

Diff since v0.1.1

Merged pull requests:

• Update CI file with documenter key and newer action versions (#55) (@stecrotti)
• Remove unnecessary compat specifications (#56) (@stecrotti)
• Add packages to list of "see also" in readme (#57) (@stecrotti)
• Add coloring to models (#58) (@stecrotti)
• Add docs for included models (#59) (@stecrotti)

v0.1.1

BeliefPropagation v0.1.1

Diff since v0.1.0

Merged pull requests:

• Remove doctest with randomness (#52) (@stecrotti)
• Update to julia 1.11 (#53) (@stecrotti)

v0.1.0

BeliefPropagation v0.1.0

Merged pull requests:

• Reduce allocations (#8) (@stecrotti)
• Add regular factor graph (#9) (@stecrotti)
• inedges and outedges are not necessary (#10) (@stecrotti)
• Make code compatible with autodiff (#11) (@stecrotti)
• Ising notebook (#12) (@stecrotti)
• Fix issues with BPFactor's (#14) (@stecrotti)
• Uniform factors return 1 instead of 1/nstates (#15) (@stecrotti)
• Add plot method for factor graph (#17) (@stecrotti)
• Detailed free energy computation (#18) (@stecrotti)
• Allow constructing a FactorGraph from an AbstractGraph (#19) (@stecrotti)
• Add docs for light free energy (#20) (@stecrotti)
• Make test functions part of the public API (#22) (@stecrotti)
• Add a simple benchmark (#23) (@stecrotti)
• Factor update by differentiating computation of za (#24) (@stecrotti)
• Add more docs (#25) (@stecrotti)
• switch to AtomicArrays (#27) (@abraunst)
• Compute $z_{ai}$ explicitly (#29) (@stecrotti)
• Small optimization (#32) (@stecrotti)
• No need for atomic vectors (#33) (@stecrotti)
• Diffres (#34) (@abraunst)
• Add compat entries for stdlib deps (#35) (@stecrotti)
• Update to julia 1.10 (#36) (@stecrotti)
• Update compat entry for IndexedGraphs (#39) (@stecrotti)
• Use IndexedFactorGraphs (#40) (@stecrotti)
• Upgrade compat for indexedgraphs (#41) (@stecrotti)
• Compute free energy not as a by-product of messages, but separately (#42) (@stecrotti)
• Fix random seed in LDPC notebook (#43) (@stecrotti)
• Improve docs (#44) (@stecrotti)
• Add K-SAT to models (#45) (@stecrotti)
• Small fix in docs (#47) (@stecrotti)
• Small fixes to docs (#48) (@stecrotti)
• Fix compat bounds for CavityTools (#49) (@stecrotti)

Closed issues:

• Compute bethe free energy as a byproduct of iterations (#1)
• Add damping and reinforcement (#2)
• Compute convergence error and stop iterations after convergence is reached (#4)
• Add multi-threading (#5)
• Write documentation (#6)
• Add regular factor graph (#7)
• Reference to unexisting LDPCStorage package (#21)
• Use AtomicVector from package (#26)
• potential bug in maxsum (#28)
• Should we compute the free energy during the iterations? (#37)
• Split FactorGraphs submodule to its own package (#38)