v0.9.0

MatrixProductBP v0.9.0

Diff since v0.8.0

v0.8.0

MatrixProductBP v0.8.0

Diff since v0.7.0

Merged pull requests:

• CompatHelper: bump compat for IndexedGraphs to 0.6, (keep existing compat) (#137) (@github-actions[bot])
• Add relative and absolute tolerances to equilibrium computations (#139) (@stecrotti)
• Optimize computation of autocorrelations (#140) (@stecrotti)
• Warning instead of error if not convergent (#141) (@stecrotti)
• Fix type instability with return nothing (#142) (@stecrotti)
• CompatHelper: bump compat for TensorTrains to 0.11, (keep existing compat) (#143) (@github-actions[bot])
• Delete slipped notebooks (#144) (@stecrotti)

v0.7.0

MatrixProductBP v0.7.0

Diff since v0.6.0

Merged pull requests:

• Update compat for tensortrains (#131) (@stecrotti)
• Follow the updates in TensorTrains.jl v0.9 -> v0.10 (#132) (@stecrotti)
• MPBP estimate of neighbor-correlations -> avg ising energy (#133) (@stecrotti)
• Add maxdist keyword for autocorrelations (#134) (@stecrotti)
• Faster method to compute the mean of a vector of Measurements (#135) (@stecrotti)
• fix small type instability (#136) (@stecrotti)

v0.6.0

MatrixProductBP v0.6.0

Diff since v0.5.0

Merged pull requests:

• Add bipartite infinite graph (#125) (@stecrotti)
• Add auto-infection to epidemic models, add tests for heterogeneous SIS (#126) (@stecrotti)
• Avoid a deepcopy when constructing MPBP from SIS (#128) (@stecrotti)
• Do not normalize messages when constructing MPBP (#129) (@stecrotti)
• Autoinfection as keyword argument (#130) (@stecrotti)

v0.5.0

MatrixProductBP v0.5.0

Diff since v0.4.0

Merged pull requests:

• Improve documentation for RecursiveBPFactor methods (#121) (@stecrotti)
• Prefer using over import (#122) (@stecrotti)
• Added the heterogeneous SIS model (#123) (@FedericoFlorio)
• New version (#124) (@stecrotti)

v0.4.0

MatrixProductBP v0.4.0

Diff since v0.3.0

Merged pull requests:

• Check that underlying BiDiGraph is symmetric (#118) (@stecrotti)
• Use flat_tt instead of uniform_tt (#119) (@stecrotti)
• CompatHelper: bump compat for CavityTools to 1, (keep existing compat) (#120) (@github-actions[bot])

v0.3.0

MatrixProductBP v0.3.0

Diff since v0.2.0

Merged pull requests:

• Use rng also for pop! (#116) (@stecrotti)
• Restrict compact of CavityTools to prevent breaking changes (#117) (@stecrotti)

v0.2.0

MatrixProductBP v0.2.0

Diff since v0.1.0

Merged pull requests:

• Errata as markdown (#107) (@stecrotti)
• More flexible TensorCast compat (#108) (@stecrotti)
• silente TensorCast.@cast warnings (#109) (@abraunst)
• Reduce allocations in generic bp (#110) (@abraunst)
• nstates(U,l) -> nstates(w::U,l) (#111) (@abraunst)
• Integerglauber (#112) (@abraunst)
• Use julia v1.10 (#113) (@stecrotti)
• Multithreaded, lighter sampling (#114) (@stecrotti)
• Add random number generator to continuous sampler (#115) (@stecrotti)

Closed issues:

• Register package (#103)

v0.1.0

MatrixProductBP v0.1.0

Merged pull requests:

• Pair observations and inference (#1) (@stecrotti)
• Re-organization as a Julia module (#2) (@stecrotti)
• add readme (#3) (@stecrotti)
• observations also at time zero (#4) (@stecrotti)
• convergence based on messages instead of beliefs (#5) (@stecrotti)
• use Tullio in SIS (#10) (@abraunst)
• Pair observations moved to only outgoing messages (#11) (@stecrotti)
• Simplifications via Tullio in SIS (#12) (@abraunst)
• replace p0 with phi (#13) (@stecrotti)
• remove two lines (#14) (@abraunst)
• Make the code for SimpleBPFactors more general (#15) (@stecrotti)
• WIP: Cavities (#17) (@abraunst)
• WIP: add codecov (#19) (@stecrotti)
• Trigger coverage when merging PR to main (#20) (@stecrotti)
• interface to SIB is in notebook (#21) (@stecrotti)
• ignore jupyter in language stats (#22) (@stecrotti)
• WIP: more than one factor simultaneously, variable number of variable states (#23) (@abraunst)
• Multithread (#25) (@abraunst)
• SIRS model, comparison with sib on SIR (#26) (@abraunst)
• Small bugfix (#27) (@stecrotti)
• avoid kron2 (#29) (@abraunst)
• Optionally print log of SVD truncations (#30) (@stecrotti)
• Equilibrium results, Erdos-Renyi, more notebooks, some small fixes (#33) (@stecrotti)
• Glauber ± J and the generalizations needed to make it work (#34) (@stecrotti)
• use @cast instead of reshape as it's nicer to read (#35) (@abraunst)
• fix multithread bug (#36) (@abraunst)
• use cavity from CavityTools (#37) (@abraunst)
• CompatHelper: add new compat entry for CavityTools at version 0.1, (keep existing compat) (#38) (@github-actions[bot])
• RFIM, plots for paper (#39) (@stecrotti)
• Plots for paper (#40) (@stecrotti)
• [WIP] Smarter free energy (#43) (@stecrotti)
• logging through logger (#44) (@abraunst)
• [WIP] Epidemics inference (#45) (@stecrotti)
• Callback enhanced with svd info (#46) (@stecrotti)
• Nicer plots, notebooks on AUC (#48) (@stecrotti)
• Large deviations + optional info string to print during progress (#49) (@stecrotti)
• import loop vectorization for Tullio usage (#50) (@abraunst)
• CompatHelper: add new compat entry for LoopVectorization at version 0.12, (keep existing compat) (#52) (@github-actions[bot])
• CompatHelper: bump compat for CavityTools to 0.2, (keep existing compat) (#53) (@github-actions[bot])
• Karate, SIRS, damping, equilibrium, TruncBondThresh, SIS vs softmargin (#54) (@stecrotti)
• CompatHelper: add new compat entry for CavityTools at version 0.2, (keep existing compat) (#55) (@github-actions[bot])
• Plots + large deviations SIS (#56) (@stecrotti)
• Effect of discarding dead epidemics, restore corrupted SIRS notebook (#57) (@stecrotti)
• Fix bug with ising ±J (#60) (@stecrotti)
• Spin glass (#61) (@stecrotti)
• "d-" for "discretized" in legend (#62) (@stecrotti)
• Asymmetric graphs (#63) (@stecrotti)
• Impose compat for LoopVectorization (#65) (@stecrotti)
• Plots with discrete CME (#67) (@stecrotti)
• Generalize marginals to MPTrain (#68) (@stecrotti)
• Equilibrium energy for infinite RRG (#69) (@stecrotti)
• Glauber correlations on infinite graph (#70) (@stecrotti)
• Update CavityTools (#71) (@stecrotti)
• Asymmetric dynamics (#72) (@stecrotti)
• Sample from MatrixProductTrain (#73) (@stecrotti)
• Use LogarithmicNumbers algebra (#74) (@stecrotti)
• Get rid of getT(::MPEM) (#75) (@stecrotti)
• [WIP] Use core code from TensorTrains.jl (#76) (@stecrotti)
• Fix small bug in arg (#77) (@stecrotti)
• Add notebook about convergence of messages (#78) (@stecrotti)
• Some stuff to reply to reviewers (#79) (@stecrotti)
• Fix issue with sampling (#80) (@stecrotti)
• Compat from julia 1.8 on (#81) (@stecrotti)
• Move recursive and test factors to main module (#84) (@stecrotti)
• Use TensorTrains v0.6 (#85) (@stecrotti)
• TensorTrains >= 0.6.1 (#87) (@stecrotti)
• Update on star graph plot (#88) (@stecrotti)
• TensorTrains v 0.7 (#89) (@stecrotti)
• Compute energy of ising model (#90) (@stecrotti)
• CompatHelper: bump compat for IndexedGraphs to 0.4, (keep existing compat) (#91) (@github-actions[bot])
• silence avx warnings (#93) (@abraunst)
• Update readme (#94) (@stecrotti)
• Optimization (#95) (@abraunst)
• Tidy up repo (#96) (@stecrotti)
• Misc (#97) (@stecrotti)
• Infected -> infectious (#98) (@stecrotti)
• CompatHelper: add new compat entry for Statistics at version 1, (keep existing compat) (#99) (@github-actions[bot])
• Generic BP adapted to work with periodic time, with test (#100) (@stecrotti)
• Add link to pnas and errata (#101) (@stecrotti)
• Add bounded compats (#104) (@stecrotti)
• Add license file, tagbot (#105) (@stecrotti)

Closed issues:

• Put pair observations only on outgoing messages (#6)
• Cavity computations (#7)
• Generalize code for SimpleBPFactors (#8)
• Replace time zero prior with observation (#9)
• Introduce MPEM1s ? (#16)
• Add codecov to github actions (#18)
• Add paralellization (#24)
• More general RecursiveBPFactors (#31)
• Warning from LoopVectorization (#58)