Merge pull request #123 from ilabcode/dev

Dev to main for next version

Project.toml

@@ -1,12 +1,14 @@
 name = "HierarchicalGaussianFiltering"
 uuid = "63d42c3e-681c-42be-892f-a47f35336a79"
-authors = ["Peter Thestrup Waade [email protected]", "Anna Hedvig Møller [email protected]", "Jacopo Comoglio [email protected]", "Christoph Mathys [email protected]\n and contributors"]
+authors = [ "Peter Thestrup Waade [email protected]",
+            "Anna Hedvig Møller [email protected]",
+            "Jacopo Comoglio [email protected]",
+            "Christoph Mathys [email protected]"]
 version = "0.3.3"
 
 [deps]
 ActionModels = "320cf53b-cc3b-4b34-9a10-0ecb113566a3"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
-LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 RecipesBase = "3cdcf5f2-1ef4-517c-9805-6587b60abb01"
 
 [compat]

README.md

@@ -91,7 +91,7 @@ plot_trajectory!(agent, ("x1", "prediction"))
 
 ````@example index
 using Distributions
-prior = Dict(("x2", "evolution_rate") => Normal(1, 0.5))
+prior = Dict(("x2", "volatility") => Normal(1, 0.5))
 
 model = fit_model(agent, prior, inputs, actions, n_iterations = 20)
 ````

docs/make.jl

@@ -61,12 +61,12 @@ makedocs(;
     ),
     pages = [
         "Introduction to Hierarchical Gaussian Filtering" => "./index.md",
-        #"Theory" => [
-        #    "./theory/genmodel.md",
-        #    "./theory/node.md",
-        #    "./theory/vape.md",
-        #    "./theory/vope.md",
-        #],
+        "Theory" => [
+            "./theory/genmodel.md",
+            "./theory/node.md",
+            "./theory/vape.md",
+            "./theory/vope.md",
+        ],
         "Using the package" => [
             "The HGF Nodes" => "./generated_markdowns/the_HGF_nodes.md",
             "Building an HGF" => "./generated_markdowns/building_an_HGF.md",

docs/src/Julia_src_files/building_an_HGF.jl

@@ -37,7 +37,7 @@ state_nodes = [
     Dict(
         "name" => "continuous_state_node",
         "type" => "continuous",
-        "evolution_rate" => -2,
+        "volatility" => -2,
         "initial_mean" => 0,
         "initial_precision" => 1,
     ),

docs/src/Julia_src_files/fitting_hgf_models.jl

@@ -47,10 +47,10 @@ using HierarchicalGaussianFiltering
 hgf_parameters = Dict(
     ("u", "category_means") => Real[0.0, 1.0],
     ("u", "input_precision") => Inf,
-    ("x2", "evolution_rate") => -2.5,
+    ("x2", "volatility") => -2.5,
     ("x2", "initial_mean") => 0,
     ("x2", "initial_precision") => 1,
-    ("x3", "evolution_rate") => -6.0,
+    ("x3", "volatility") => -6.0,
     ("x3", "initial_mean") => 1,
     ("x3", "initial_precision") => 1,
     ("x1", "x2", "value_coupling") => 1.0,
@@ -95,12 +95,12 @@ fixed_parameters = Dict(
     ("x3", "initial_precision") => 1,
     ("x1", "x2", "value_coupling") => 1.0,
     ("x2", "x3", "volatility_coupling") => 1.0,
-    ("x3", "evolution_rate") => -6.0,
+    ("x3", "volatility") => -6.0,
 );
 
 # As you can read from the fixed parameters, the evolution rate of x2 is not configured. We set the prior for the x2 evolution rate:
 using Distributions
-param_priors = Dict(("x2", "evolution_rate") => Normal(-3.0, 0.5));
+param_priors = Dict(("x2", "volatility") => Normal(-3.0, 0.5));
 
 # We can fit the evolution rate by inputting the variables:
 

docs/src/Julia_src_files/index.jl

@@ -53,7 +53,7 @@ plot_trajectory!(agent, ("x1", "prediction"))
 # ### Fitting parameters
 
 using Distributions
-prior = Dict(("x2", "evolution_rate") => Normal(1, 0.5))
+prior = Dict(("x2", "volatility") => Normal(1, 0.5))
 
 model = fit_model(agent, prior, inputs, actions, n_iterations = 20)
 

docs/src/Julia_src_files/premade_HGF.jl

@@ -9,7 +9,6 @@
 # - [Categorical 3-level HGF](#Categorical-3-level-HGF)
 # - [Categorical 3-level state transition HGF](#Categorical-3-level-state-transition)
 
-# ![continuous 2-level graph](../images/HGF_structures/all_models.png)
 
 # #Load data for examples
 
@@ -40,13 +39,19 @@ inputs_binary = CSV.read(hgf_path_binary * "classic_binary_inputs.csv", DataFram
 
 # ## Continuous 2-level HGF
 
+
+
+# ![continuous 2-level graph](../images/HGF_structures/continuous.svg)
+
+
 # The continuous 2-level HGF is structured with following nodes:
 
 # - input node: continuous 
 # - state nodes: 
 #   - 1st level: continuous (value coupling to input node)
 #   - 2nd level: continous (volatility coupling to 1st level)
 
+
 #Create HGF and Agent
 continuous_2_level = premade_hgf("continuous_2level");
 agent_continuous_2_level =
@@ -93,6 +98,9 @@ plot_trajectory(
 
 # ## Binary 2-level HGF
 
+# ![continuous 2-level graph](../images/HGF_structures/binary.svg)
+
+
 # - input node: binary
 # - state nodes:
 #   - 1st level: binary (value coupling to input node)
@@ -143,6 +151,9 @@ plot_trajectory(agent_binary_3_level, ("x3", "posterior"))
 
 # ## Categorical 3-level HGF
 
+# ![continuous 2-level graph](../images/HGF_structures/categorical.svg)
+
+
 # The categorical 3-level HGF model takes an input from one of m categories and learns the probability of a category appearing.
 
 # - input node: categorical

docs/src/Julia_src_files/premade_models.jl

@@ -75,7 +75,7 @@ set_parameters!(agent, ("x3", "initial_precision"), 0.4)
 # Set multiple parameter values
 set_parameters!(
     agent,
-    Dict(("x3", "initial_precision") => 1, ("x3", "evolution_rate") => 0),
+    Dict(("x3", "initial_precision") => 1, ("x3", "volatility") => 0),
 )
 
 

docs/src/Julia_src_files/the_HGF_nodes.jl

@@ -2,67 +2,13 @@
 
 # In this section we will cover the types of nodes, their parameters and the rules for structuring your own HGF. 
 
-
 # ### Overview
 
-#   - [The Node Types In an HGF and their states](#The-Node-Types-In-an-HGF)
-#   - [Building principles ](#Building-principles )
-
-# ### The Node Types In an HGF
-
-# We have six types of nodes in the HGF: binary (state node and input node), categorical (state node and input node), and continuous (state node and input node).
-
-# If a node's parameters are configured with a default value, they are stated below as well. 
-
-# We provide an overview of the states in each of the nodes. 
-
-# 1. [Binary Nodes](#Binary-Nodes)
-#     1. [State Node: States and parameters](#The-states-of-binary-state-nodes-and-parameters)
-#     2. [Input Node: States and parameters](#The-states-of-binary-input-nodes-and-parameters)
-
-# 3. [Categorical Nodes](#Categorical-Nodes)
-#    1. [State Node: States and parameters](#The-states-of-Categorical-state-nodes-and-parameters)
-#    2. [Input Node: States and parameters](#The-states-of-Categorical-Input-nodes-and-parameters)
-
-# 4. [Continuous Nodes](#Continuous-Nodes)
-#    1. [State Node: States and parameters](#The-states-of-Continuous-state-nodes-and-parameters)
-#    2. [Input Node: States and parameters](#The-states-of-Continuous-Input-nodes-and-parameters)
-
-
-# ## Binary Nodes
-
-# ### The states of binary state nodes
-
-# - posterior_mean
-# - posterior_precision
-# - value\_prediction\_error
-# - prediction_mean
-# - prediction_precision
-
-# ### Parameters
-
-# - Value coupling
-
-
-# ### The states of binary input nodes and parameters
+# ![states and parameters](../images/states_nodes/all_states.svg)
 
-# - input value
-# - value prediction error
-
-# ### Parameters
-
-# - Category means (default is [0,1])
-# - Input precision (default is  infinite input precision)
-
-
-# ## Categorical Nodes
-
-# ### The states of Categorical state nodes and parameters
-
-# - posterior
-# - value prediction error
-# - prediction
+# ## Building principles
 
+# The following rules apply for connecting nodes, when customizing your own HGF structure:
 # ### Parameters
 
 # - no parameters in the categorical state node
@@ -109,20 +55,18 @@
 
 # ### Parameters
 
-# - evolution rate (default is 0)
+# - input noise (default is 0)
 # - value coupling 
 # - volatility coupling
-
-
-
-# ## Building principles 
 
 # ### Binary state node rules:
 
 # - Can only have exactly one value parent
 # - Can only have excatly one value child
 # - Can only have a contionus state node as value parent
 
+# ![states and parameters](../images/states_nodes/binary_nodes.svg)
+
 # ### continuous state node rules:
 
 # - Can’t have binary input node as child
@@ -132,8 +76,12 @@
 # - Can’t have the same value parent as volatility parent
 # - Can’t have the same value child as volatility child
 
+# ![states and parameters](../images/states_nodes/continuous_nodes.svg)
+
 # ### Categorical state node rules:
 
 # - Can only have exactly one value child
 # - Can only have categorical input node as child
 # - Can only have binary state node as parents
+
+# ![states and parameters](../images/states_nodes/categorical_nodes.svg)

docs/src/Julia_src_files/utility_functions.jl

@@ -34,7 +34,7 @@ get_parameters(agent)
 # ERROR WITH THIS get_parameters(agent, ("x2", "x3", "volatility_coupling"))
 
 # getting multiple parameters specify them in a vector
-get_parameters(agent, [("x3", "evolution_rate"), ("x3", "initial_precision")])
+get_parameters(agent, [("x3", "volatility"), ("x3", "initial_precision")])
 
 
 # ### Getting States
@@ -46,7 +46,7 @@ get_states(agent)
 get_states(agent, ("x2", "posterior_precision"))
 
 #getting multiple states
-get_states(agent, [("x2", "posterior_precision"), ("x2", "auxiliary_prediction_precision")])
+get_states(agent, [("x2", "posterior_precision"), ("x2", "volatility_weighted_prediction_precision")])
 
 
 # ### Setting Parameters
@@ -61,10 +61,10 @@ agent_parameter = Dict("sigmoid_action_precision" => 3)
 hgf_parameters = Dict(
     ("u", "category_means") => Real[0.0, 1.0],
     ("u", "input_precision") => Inf,
-    ("x2", "evolution_rate") => -2.5,
+    ("x2", "volatility") => -2.5,
     ("x2", "initial_mean") => 0,
     ("x2", "initial_precision") => 1,
-    ("x3", "evolution_rate") => -6.0,
+    ("x3", "volatility") => -6.0,
     ("x3", "initial_mean") => 1,
     ("x3", "initial_precision") => 1,
     ("x1", "x2", "value_coupling") => 1.0,