Recent Releases of forneylab.jl

forneylab.jl - v0.12.0

ForneyLab v0.12.0

Diff since v0.11.4

Closed issues: - Use PDMats.jl to work with positive-definite matrices (#30) - Cryptic error messages (#27) - Write a "how to build a node" demo (#29) - Improve error message for algo contruction on model with dangling edges (#70) - How to implement boolean operations/factors for binary random variables, or more generally conditional probability tables? (#100) - Probability density estimation for marginals parameterised with SampleList (#137) - ERROR: LoadError: KeyError: key nothing not found (#146) - Error: No applicable marginal update rule for GaussianMeanVariance node with inbound types: Message{GaussianWeightedMeanPrecision}, Message{GaussianMeanVariance} (#194) - Multiplication of two categorical variables (#197) - Incorrect outbound message update rules for GaussianMeanVariance (#198) - Where are parameters stored in ForneyLab.Bernoulli? (#199) - ForneyLab.step! must be explicitly imported (#200)

Merged pull requests: - Refactor rule matching (#188) (@ThijsvdLaar) - Natural parameter distributions (#189) (@ThijsvdLaar) - Conjugate variational Inference (#192) (@ThijsvdLaar) - Rename (#193) (@ThijsvdLaar) - Update citation file (#195) (@ThijsvdLaar) - Fix gaussian update (#203) (@ThijsvdLaar) - Improve error messages (#204) (@ThijsvdLaar) - Update compat (#205) (@ThijsvdLaar) - Write And node demo (#206) (@ThijsvdLaar) - Prepare for release 0.12.0 (#208) (@ThijsvdLaar)

- Julia
Published by github-actions[bot] about 4 years ago

forneylab.jl - v0.11.4

ForneyLab v0.11.4

Diff since v0.11.3

Closed issues: - Algorithm construction bug (Nonlinear node) (#166) - Question: Can ForneyLab do parameter learning and graph structure learning? (#176) - code generation for repetitive graph structure (#179) - Belief Propagation (#185)

Merged pull requests: - Constraints (#178) (@ThijsvdLaar) - Transition Mixture Node (#180) (@ThijsvdLaar) - Nonlinear fix (#181) (@albertpod) - Patch for numerical stability of transition mixture (#184) (@ThijsvdLaar) - Refactor Nonlinear dimensionality specification (#187) (@ThijsvdLaar) - Bump FL to version 0.11.4 (#190) (@ThijsvdLaar) - Update compat for automated release (#191) (@ThijsvdLaar)

- Julia
Published by github-actions[bot] over 4 years ago

forneylab.jl - v0.11.3

ForneyLab v0.11.3

Diff since v0.11.2

Closed issues: - Improve stability when inverting positive definite matrices (#86) - Turing, SampledSignals and ForneyLab compatibility issues (#129) - After last updates, ForneyLab cannot build message passing algorithm. (#132) - Scheduler fails for posterior factors surrounding Addition node (#142) - Confidence Intervals on parameters (#148) - Free energy code generation using custom parameterized factor nodes (#149) - Naive beginner's question - categorical variable input (#152) - Stability of the log(det()) function (#153) - Resulting Gaussians improper (#156) - messagePassingAlgorithm not found (#159) - MethodError: no method matching length(::Variable) (#161) - KeyError: key :log_pdf not found (#162)

Merged pull requests: - compat(): Fix for compatibility issue #129 (#130) (@bvdmitri) - Make FactorNode and Function prod symmetric (#133) (@semihakbayrak) - Introduce strategy for stable Cholesky factorizations (#134) (@ThijsvdLaar) - CompatHelper: add new compat entry for "PositiveFactorizations" at version "0.2" (#138) (@github-actions[bot]) - Refactor user interface for EP update specification (#139) (@ThijsvdLaar) - Multivariate nonlinear function of non Gaussian inputs (#141) (@semihakbayrak) - fix bug - removePrefix() (#150) (@bartvanerp) - A more universal entropy calculation method for sampling based inference (#151) (@semihakbayrak) - Stability - improve log(det()) function (#154) (@bartvanerp) - Update missed log(det()) functions (#158) (@bartvanerp) - bump version (#160) (@ThijsvdLaar) - fix(): Fix for #162 (#163) (@bvdmitri) - update(): Bump compat entries to be compatible with ReactiveMP.jl (#168) (@bvdmitri) - update(): Update Project.toml to satisfy Julia General registry auto-merge requirements (#169) (@bvdmitri) - update(): Add TagBot github workflow for automatic releases and tags (#170) (@bvdmitri) - Fix probit average energy (#171) (@bartvanerp) - Migrate from TravisCI to GitHub Actions (#172) (@bvdmitri) - Change LICENSE.md to MIT (#173) (@bvdmitri)

- Julia
Published by github-actions[bot] almost 5 years ago

forneylab.jl - ForneyLab.jl v0.11.2

  • Extended message passing rules around Equality node
  • New demo with estimation of player skill ratings
  • Bug fixes and improvements

- Julia
Published by ivan-bocharov over 5 years ago

forneylab.jl - ForneyLab.jl v0.11.1

  • Unified message passing algorithm construction
  • New approximation method for Nonlinear nodes
  • New sampling-based rules for Gamma, Beta, and Dirichlet distributions
  • Automated documentation builds

- Julia
Published by ivan-bocharov almost 6 years ago

forneylab.jl - ForneyLab.jl v0.11.0

  • Implement support for sequential Monte-Carlo
  • Implement support for local variational methods
  • Decouple algorithm assembly from code generation
  • Extend Nonlinear node implementation to multiple arguments
  • Refactor recognition factorization handling

- Julia
Published by ThijsvdLaar almost 6 years ago

forneylab.jl - ForneyLab.jl v0.10.0

  • New implementation of Nonlinear node using unscented transform
  • Changes in update rule naming scheme (Void -> V is substituted by Nothing -> N)
  • Code style guide

- Julia
Published by ivan-bocharov almost 7 years ago

forneylab.jl - ForneyLab.jl v0.9.2

  • Poisson node implementation
  • New documentation system
  • Minor fixes (#14, #15, #32, #33)

- Julia
Published by ivan-bocharov about 7 years ago

forneylab.jl - ForneyLab.jl version 0.9.1

This version incorporates some minor changes that enable more accurate backward propagation through nonlinear nodes.

- Julia
Published by ivan-bocharov over 7 years ago

forneylab.jl - ForneyLab.jl v0.9.0

This release migrates ForneyLab.jl to Julia v0.7 and v1.0.

- Julia
Published by marcocox almost 8 years ago