Albert

Following our [discussion](https://github.com/biaslab/ReactiveMP.jl/pull/132#discussion_r896459954), we need to change the update rule for `ruleVBPoissonOut(marg_out::Any, marg_l::Distribution{Univariate, Gamma})` as in `ReactiveMP.jl` (see [out.jl](https://github.com/biaslab/ReactiveMP.jl/blob/4de9106eeac3b8d613bb58d01a7698d7d4223b34/src/rules/poisson/out.jl#L5))

This PR adds a rule for multiplication node: ```julia ruleSPMultiplicationAGPN(msg_out::Message{F, Multivariate}, msg_in1::Message{PointMass, Multivariate}, msg_a::Nothing) ``` ``` ^ | a ~ Univariate out ~ Multivariate -->[x]

I am trying to build the inference algorithm in the coupled random walk model:  We assume the following factorization: `q = PosteriorFactorization(x_0, x, γ_x,...

It would be nice to support the `check_args` argument for matrix/multivariate distributions, similarly to univariate distributions. For Inverse-Gamma distribution we can write: `InverseGamma(-1, 1, check_args=false)` While for Inverse-Wishart, `Distributions.jl `throws...

Complete optimization procedure for multivariate and matrix variate random variables.

1

The current implementation of [PointMassFormConstraint](https://github.com/biaslab/ReactiveMP.jl/blob/master/src/constraints/form/form_point_mass.jl) aka EM-like MP doesn't support multivariate or matrix variate distributions.

Importance sampling for GammaMixture support.

4

We have (tw @bvdmitri ) spotted an issue with running an important sampling strategy around the GammaMixture node. We proposed this strategy n [this paper](https://ieeexplore.ieee.org/document/9596329). The API for the form...

Given the recent (and old) updates of `ReactiveMP.jl`, the current README does not fully represent what our framework is capable of. The framework works beyond only conjugate SSMs, so we...

Add SGCV node

1

We should add the switching gaussian controlled variance `SGCV` node.

Add Delta node with Extended and Unscented transform

5

This PR implements two types of approximation of messages around the Delta node, i.e. DeltaExtended (ET) and DeltaUnscented (UT). Both ET and UT approximate the messages with a Gaussian. To...

[Basic Kalman Smoothing with LSTM driven dynamics](https://github.com/biaslab/ReactiveMP.jl/blob/master/demo/broken/Basic%20Kalman%20Smoothing%20with%20LSTM%20driven%20dynamics.ipynb) doesn't work. It breaks with the following error: ```julia MethodError: no method matching Flux.LSTMCell(::Matrix{Float64}, ::Matrix{Float64}, ::Vector{Float64}, ::Vector{Float64}, ::Vector{Float64}) Closest candidates are: Flux.LSTMCell(::A, ::A,...