Tor Erlend Fjelde torfjelde

## prophet-using-Turing.jl
# Partial implementation of linear-trend model described in [1] using Turing.jl.
#
# # References
# [1] Taylor, S. J., & Letham, B., Forecasting at scale, PeerJ Preprints, 5(), 3190–2 (2017).  http://dx.doi.org/10.7287/peerj.preprints.3190v2

using DataFrames, CSV
using Turing
using MCMCChain, Plots, StatsPlots

pyplot()

## advi.png

      
              2 files
            
          
              0 forks
            
          
              5 comments
            
          
              2 stars
            
          
                torfjelde
                / advi.png
            
            
              Last active
              May 17, 2019 11:17
            
              
                An example of how one can implement Automatic Derivative Variational Inference (VI) for Turing.jl.
              
          
## turing_plot.jl
using MacroTools
using Turing

using LightGraphs, MetaGraphs

# Expressions
ex1 = quote
    m(x) = begin
        # Assumptions
        σ ~ InverseGamma(2,3)

## bijectors_interface.jl
using Distributions, Bijectors
using ForwardDiff
using Tracker

using Turing

import Random: AbstractRNG
import Distributions: logpdf, rand, rand!, _rand!, _logpdf


## ADVI-MNIST.jl
{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ADVI on MNIST"
   ]
  },
  {

## FSSD_gof_pushforward.jl
julia> using Bijectors

julia> using KernelGoodnessOfFit

julia> using Random; Random.seed!(123);

julia> using ForwardDiff

julia> function Distributions.gradlogpdf(d::ContinuousUnivariateDistribution, x::Real)
           ForwardDiff.derivative(z -> logpdf(d, z), x)

## turing-condition.jl
julia> using Turing

julia> function condition(
           model::Turing.Model,
           latent::NamedTuple;
           sampler = DynamicPPL.SampleFromPrior()
       )
           vi = Turing.VarInfo(model)
           md = vi.metadata

## bijector-macro.jl
"""
    @bijector function f(b::Bijector, x) ... end

Takes the method `forward` and uses it to define both `transform`
and `logabsdetjac`, while ensuring that any shared computation is
taken advantage of in `forward`.
"""
macro bijector(expr)
    def = MacroTools.splitdef(expr)
    body = def[:body]

## manual-sample.jl
julia> using Random

julia> using Turing

julia> rng = MersenneTwister(42);

julia> @model function demo(x)
           s ~ InverseGamma(2, 3)
           m ~ Normal(0, √s)
           for i in eachindex(x)

## logging-while-sampling.jl
using Turing
using TensorBoardLogger, Logging
using OnlineStats      # used to compute different statistics on-the-fly
using StatsBase        # Provides us with the `Histogram` which is supported by `TensorBoardLogger.jl`
using LinearAlgebra
using DataStructures   # will use a `CircularBuffer` to only keep track of some `n` last samples

struct TBCallback
    logger::TBLogger
end
	# Partial implementation of linear-trend model described in [1] using Turing.jl.
	#
	# # References
	# [1] Taylor, S. J., & Letham, B., Forecasting at scale, PeerJ Preprints, 5(), 3190–2 (2017). http://dx.doi.org/10.7287/peerj.preprints.3190v2

	using DataFrames, CSV
	using Turing
	using MCMCChain, Plots, StatsPlots

	pyplot()
	using MacroTools
	using Turing

	using LightGraphs, MetaGraphs

	# Expressions
	ex1 = quote
	m(x) = begin
	# Assumptions
	σ ~ InverseGamma(2,3)
	using Distributions, Bijectors
	using ForwardDiff
	using Tracker

	using Turing

	import Random: AbstractRNG
	import Distributions: logpdf, rand, rand!, _rand!, _logpdf
	{
	"cells": [
	{
	"cell_type": "markdown",
	"metadata": {},
	"source": [
	"# ADVI on MNIST"
	]
	},
	{
	julia> using Bijectors

	julia> using KernelGoodnessOfFit

	julia> using Random; Random.seed!(123);

	julia> using ForwardDiff

	julia> function Distributions.gradlogpdf(d::ContinuousUnivariateDistribution, x::Real)
	ForwardDiff.derivative(z -> logpdf(d, z), x)
	julia> using Turing

	julia> function condition(
	model::Turing.Model,
	latent::NamedTuple;
	sampler = DynamicPPL.SampleFromPrior()
	)
	vi = Turing.VarInfo(model)
	md = vi.metadata
	"""
	@bijector function f(b::Bijector, x) ... end

	Takes the method `forward` and uses it to define both `transform`
	and `logabsdetjac`, while ensuring that any shared computation is
	taken advantage of in `forward`.
	"""
	macro bijector(expr)
	def = MacroTools.splitdef(expr)
	body = def[:body]
	julia> using Random

	julia> using Turing

	julia> rng = MersenneTwister(42);

	julia> @model function demo(x)
	s ~ InverseGamma(2, 3)
	m ~ Normal(0, √s)
	for i in eachindex(x)
	using Turing
	using TensorBoardLogger, Logging
	using OnlineStats # used to compute different statistics on-the-fly
	using StatsBase # Provides us with the `Histogram` which is supported by `TensorBoardLogger.jl`
	using LinearAlgebra
	using DataStructures # will use a `CircularBuffer` to only keep track of some `n` last samples

	struct TBCallback
	logger::TBLogger
	end