{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Manifold MCMC methods for diffusions: *FitzHugh-Nagumo* model example\n",
"\n",
"This [Jupyter notebook](https://jupyter.org/) accompanies the paper [*Manifold MCMC methods for Bayesian inference in a wide class of diffusion models*](https://arxiv.org/abs/1912.02982), providing a complete runnable example of applying the method described in the paper to perform inference in an example hypoelliptic diffusion model.\n",
"\n",
"## Setup\n",
"\n",
"We first check if the notebook is being run on [Binder](https://mybinder.org/) or [Google Colab](https://colab.research.google.com/) and if so install the `sde` package and the other dependencies using `pip`."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"ON_BINDER = 'BINDER_SERVICE_HOST' in os.environ\n",
"\n",
"try:\n",
" import google.colab\n",
" ON_COLAB = True\n",
"except:\n",
" ON_COLAB = False\n",
"\n",
"if ON_COLAB:\n",
" !pip install git+https://github.com/thiery-lab/manifold-mcmc-for-diffusions.git#egg=sde[notebook]\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We now import the modules we will use to simulate from the model and perform inference"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"import mici\n",
"import sde.mici_extensions as mici_extensions\n",
"import symnum\n",
"import symnum.diffops.symbolic as diffops\n",
"import symnum.numpy as snp\n",
"import numpy as onp\n",
"import jax\n",
"from jax import lax, config, numpy as jnp\n",
"import matplotlib.pyplot as plt\n",
"import arviz\n",
"import corner\n",
"\n",
"config.update('jax_enable_x64', True)\n",
"config.update('jax_platform_name', 'cpu')\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We also set a dictionary of style parameters to use with Matplotlib plots"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"plot_style = {\n",
" 'mathtext.fontset': 'cm',\n",
" 'font.family': 'serif',\n",
" 'axes.titlesize': 10,\n",
" 'axes.labelsize': 10,\n",
" 'xtick.labelsize': 6,\n",
" 'ytick.labelsize': 6,\n",
" 'legend.fontsize': 8,\n",
" 'legend.frameon': False,\n",
" 'axes.linewidth': 0.5,\n",
" 'lines.linewidth': 0.5,\n",
" 'axes.labelpad': 2.,\n",
" 'figure.dpi': 150,\n",
"}\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Diffusion model\n",
"\n",
"As an illustration we will consider the hypoelliptic diffusion defined by the system of *stochastic differential equations* (SDEs)\n",
"\n",
"$$\n",
" \\underbrace{\\begin{bmatrix} \\mathrm{d} \\mathsf{x}_0(\\tau) \\\\ \\mathrm{d} \\mathsf{x}_1(\\tau) \\end{bmatrix}}_{\\mathrm{d}\\mathsf{x}(\\tau)} = \n",
" \\underbrace{\\begin{bmatrix}\n",
" \\frac{1}{\\epsilon} (\\mathsf{x}_0(\\tau) - \\mathsf{x}_0(\\tau)^3 - \\mathsf{x}_1(\\tau)) \\\\\n",
" \\gamma \\mathsf{x}_0(\\tau) - \\mathsf{x}_1(\\tau) + \\beta\n",
" \\end{bmatrix}}_{a(\\mathsf{x}(\\tau),\\mathsf{z})} \\mathrm{d} \\tau + \n",
" \\underbrace{\\begin{bmatrix} 0 \\\\ \\sigma \\end{bmatrix}}_{B(\\mathsf{x}(\\tau),\\mathsf{z})} \\mathrm{d} \\mathsf{w}(\\tau)\n",
"$$\n",
"with $\\mathsf{x}$ the $\\mathcal{X} = \\mathbb{R}^2$-valued diffusion process of interest, $\\mathsf{w}$ a univariate Wiener process and $\\mathsf{z} = [\\sigma;\\epsilon;\\gamma;\\beta] \\in \\mathcal{Z} =\\mathbb{R}_{>0} \\times \\mathbb{R}_{>0} \\times \\mathbb{R} \\times \\mathbb{R}$ the model parameters. \n",
"\n",
"This SDE system corresponds to a stochastic variant of the [Fitzhugh-Nagumo model](http://www.scholarpedia.org/article/FitzHugh-Nagumo_model), a simplified description of actional potential generation within a neuronal axon.\n",
"\n",
"We will use [SymNum](https://github.com/matt-graham/symnum) to symbolically define the drift $a$ and diffusion coefficient $B$ functions for the model in terms of the current state $\\mathsf{x}$ and parameters $\\mathsf{z} = [\\sigma;\\epsilon;\\gamma;\\beta]$. This will later allow us to automatically construct a function to numerical integrate the SDE system."
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"dim_x = 2\n",
"dim_w = 1\n",
"dim_z = 4\n",
"\n",
"def drift_func(x, z):\n",
" σ, ε, γ, β = z\n",
" return snp.array([(x[0] - x[0]**3 - x[1]) / ε, γ * x[0] - x[1] + β])\n",
"\n",
"def diff_coeff(x, z):\n",
" σ, ε, γ, β = z\n",
" return snp.array([[0], [σ]])\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Time discretisation\n",
"\n",
"As in general exact simulation of the diffusion models of interest will be intractable, we define an approximate discrete time model based on numerical integration of the SDEs. Various numerical schemes for integrating SDE systems are available with varying convergence properties and implementational complexity - see for example [*Numerical Solutions of Stochastic Differential Equations* (Kloden and Platen, 1992)](https://books.google.com.sg/books/about/Numerical_Solution_of_Stochastic_Differe.html?id=7bkZAQAAIAAJ&source=kp_book_description&redir_esc=y) for an in-depth survey.\n",
"\n",
"The simplest and most common scheme is the *Euler-Maruyama* method (corresponding to a strong-order 0.5 Taylor approximation), which for a small time step $\\delta > 0$ can be defined by a *forward operator* $f_{\\delta} : \\mathcal{Z} \\times \\mathcal{X} \\times \\mathcal{V} \\to \\mathcal{X}$\n",
"\n",
"$$\n",
" f_\\delta(z, x, v) = \n",
" x + \\delta {a}(x, z) + \n",
" \\delta^\\frac{1}{2} B(x,z) v\n",
"$$\n",
"\n",
"where $v \\in \\mathcal{V}$ is a vector of independent standard normal random variates of dimension equal to that of the Wiener process (here one).\n",
"\n",
"The corresponding single step update can be defined using SymNum as:"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"def euler_maruyama_step(z, x, v, δ):\n",
" return x + δ * drift_func(x, z) + δ**0.5 * diff_coeff(x, z) @ v"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"More accurate approximations can be derived by using higher-order terms from the stochastic Taylor expansion of the SDE system. For example for a SDE model with *additive noise*, i.e. a diffusion coefficient $B$ which is independent of the state $B(x, z) = B(z)$, a *strong order 1.5 Taylor scheme* can be defined by the forward operator\n",
"\n",
"$$\n",
" f_\\delta(z, x, [v_1; v_2]) = \n",
" x + \\delta a(x, z) + \n",
" \\frac{\\delta^2}{2} \\partial_0 a(x,z) a(x,z) + \n",
" \\frac{\\delta^2}{4}[(\\mathrm{tr}(\\partial^2_1 a_i(x, z) B(z) B(z)^{\\rm T}))_{i=0}^{\\mathtt{X}-1}] +\n",
" \\delta^{\\frac{1}{2}} B(z) v_0 +\n",
" \\frac{\\delta^{\\frac{3}{2}}}{2} \\partial_1 a(x,z) B(z) (v_0 + v_1 / \\sqrt{3})\n",
"$$\n",
"\n",
"with both $v_1$ and $v_2$ having the dimension of the Wiener process and so the vector $v = [v_0; v_1]$ twice the dimension of the Wiener process (therefore of dimension 2 here). This can be implemented using SymNum as follows"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"def strong_order_1p5_step(z, x, v, δ):\n",
" a = drift_func(x, z)\n",
" da_dx = diffops.jacobian(drift_func)(x, z)\n",
" B = diff_coeff(x, z)\n",
" dim_noise = B.shape[1]\n",
" d2a_dx2_BB = diffops.matrix_hessian_product(drift_func)(x, z)(B @ B.T)\n",
" v_1, v_2 = v[:dim_noise], v[dim_noise:]\n",
" return (\n",
" x + δ * a + (δ**2 / 2) * da_dx @ a + (δ**2 / 4) * d2a_dx2_BB + \n",
" δ**0.5 * B @ v_1 + (δ**1.5 / 2) * da_dx @ B @ (v_1 + v_2 / snp.sqrt(3)))\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We can use these symbolically defined single step updates to define corresponding numerical functions which take NumPy arrays as inputs using SymNum's `numpify`_func function. As well as the function to be transformed, the `numpify_func` function requires the shape (dimensions) of all arguments to be specified. It also optionally allows specifying the module to use for the NumPy API calls with here we using the `jax.numpy` module from [JAX](https://github.com/google/jax) as this will allow us to later automatically construct efficient derivative functions for inference. Below we define a forward operator function using the strong order 1.5 step however we can instead use the Euler-Maruyma discretisation simply by setting the `use_euler_maruyama` flag to `True`."
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"use_euler_maruyama = False\n",
"if use_euler_maruyama:\n",
" forward_func = euler_maruyama_step\n",
" dim_v = dim_w\n",
"else:\n",
" forward_func = strong_order_1p5_step\n",
" dim_v = 2 * dim_w\n",
"forward_func = symnum.numpify_func(\n",
" forward_func, (dim_z,), (dim_x,), (dim_v,), None, numpy_module=jnp\n",
")\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Given a forward operator we can generate (approximate) samples of the state process at a series of discrete times. Here we assume that we use a fixed time increment $\\delta > 0$ for all integrator steps and denote $\\mathsf{x}_{\\texttt{s}}$ as the approximation to $\\mathsf{x}(\\mathtt{s}\\delta)$.\n",
"\n",
"## Observation model\n",
"\n",
"As in the paper we assume the simple case that the diffusion process is discretely observed at $\\texttt{T}$ equally spaced times $\\tau_\\texttt{t} = \\texttt{t}\\Delta~~\\forall \\texttt{t}\\in 1{:}\\texttt{T}$. We use a fixed number of steps $\\texttt{S}$ per interobservation interval with $\\delta = \\frac{\\Delta}{\\texttt{S}}$ so that the state at the $\\texttt{t}$th observation time is $\\mathsf{x}_{\\texttt{St}}$ and the whole sequence of states to be simulated is $\\mathsf{x}_{1{:}\\mathtt{ST}}$.\n",
"\n",
"We assume the $\\mathtt{Y} = 1$ dimensional observations $\\mathsf{y}_{1{:}\\mathtt{T}}$ correspond to direct observation of the first state component i.e. $\\mathsf{y}_\\texttt{t} = h_\\mathtt{t}(\\mathsf{x}) = \\mathsf{x}_0 ~~\\forall \\mathtt{t} \\in 1{:}\\mathtt{T}$."
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"def obs_func(x_seq):\n",
" return x_seq[..., 0:1]\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Generative model"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"As described in the paper we use a non-centered parameterisation of the generative model for the parameters $\\mathsf{z}$, time-discretised diffusion $\\mathsf{x}_{0{:}\\mathtt{ST}}$ and observations $\\mathsf{y}_{1{:}\\mathtt{T}}$\n",
"\n",
"We use priors $\\mathsf{x}_{0} \\sim \\mathcal{N}([-0.5;-0.5], \\mathbb{I}_2)$, $\\log{\\sigma} \\sim \\mathcal{N}\\left(-1, 0.5^2\\right)$, $\\log{\\epsilon} \\sim \\mathcal{N}\\left(-2, 0.5^2\\right)$, ${\\gamma} \\sim \\mathcal{N}\\left(1, 0.5^2\\right)$ and ${\\beta} \\sim \\mathcal{N}\\left(1, 0.5^2\\right)$ which were roughly tuned so that with high probability state sequences $\\mathsf{x}_{1{:}\\mathtt{ST}}$ generated from the prior exhibited stable spiking dynamics and such that $\\sigma$ and $\\epsilon$ obey their positivity constraints. \n",
"\n",
"We reparameterise the parameters $\\mathsf{z}$ and initial state $\\mathsf{x}_0$ in terms of vectors of standard normal variates, respectively $\\mathsf{u}$ and $\\mathsf{v}_0$, with the parameter and initial state generator functions then set to $g_{\\mathsf{z}}(u) = [\\exp(0.5 u_0 -1); \\exp(0.5 u_1 - 2); 0.5 u_2 + 1; 0.5 u_3 + 1]$ and $g_{\\mathsf{x}_0}(v_0, z) = [v_{0,0} -0.5; v_{0,1} - 0.5]$ with input distributions $\\tilde{\\mu} = \\mathcal{N}(0,\\mathbb{I}_4)$ and $\\tilde{\\nu} = \\mathcal{N}(0,\\mathbb{I}_2)$. We can implement these generator functions in Python using JAX NumPy API functions (to allow us to later algorithmically differentiate through the generative model) as follows. "
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [],
"source": [
"def generate_z(u):\n",
" \"\"\"Generate parameters from prior given an standard normal vector.\"\"\"\n",
" return jnp.array([\n",
" jnp.exp(0.5 * u[0] - 1), # σ\n",
" jnp.exp(0.5 * u[1] - 2), # ϵ\n",
" 0.5 * u[2] + 1, # γ\n",
" 0.5 * u[3] + 1, # β\n",
" ])\n",
"\n",
"def generate_x_0(z, v_0):\n",
" \"\"\"Generate an initial state from prior given a standard normal vector.\"\"\"\n",
" return jnp.array([-0.5, -0.5]) + v_0\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The overall joint generative model for $\\mathsf{z}$, $\\mathsf{x}_{0{:}\\mathtt{ST}}$ and $\\mathsf{y}_{1{:}\\mathtt{T}}$ in terms of the independent and standard normal variates $\\mathsf{u}$ and $\\mathsf{v}_{0{:}\\mathtt{ST}}$ can then be summarised\n",
"\n",
"$$\n",
"\\begin{aligned}\n",
" \\mathsf{u} &\\sim \\mathcal{N}(0, \\mathbb{I}_4) \\\\\n",
" \\mathsf{v}_\\texttt{s} &\\sim \\mathcal{N}(0, \\mathbb{I}_2) \\quad &\\forall \\mathtt{s} \\in 0{:}\\mathtt{ST}\\\\\n",
" \\mathsf{z} &= g_{\\mathsf{z}}(\\mathsf{u})\\\\\n",
" \\mathsf{x}_0 &= g_{\\mathsf{x}_0}(\\mathsf{v}_0, \\mathsf{z}) \\\\\n",
" \\mathsf{x}_{\\mathtt{s}+1} &= f_{\\delta}(\\mathsf{z}, \\mathsf{x}_{\\mathtt{s}}, \\mathsf{v}_{\\mathtt{s}})\n",
" \\quad &\\forall \\mathtt{s} \\in 1{:}\\mathtt{ST}\\\\\n",
" \\mathsf{y}_{\\mathtt{t}} &= h_{\\mathtt{t}}(\\mathsf{x}_{\\mathtt{St}})\n",
" \\quad &\\forall \\mathtt{t} \\in 1{:}\\mathtt{T}\n",
"\\end{aligned}\n",
"$$\n",
"\n",
"We collect all of the latent variables in to a $6 + 2\\texttt{ST}$ dimensional flat vector $\\mathsf{q} := [\\mathsf{u}; \\mathsf{v}_{0{:}\\mathtt{ST}}]$, with all components of $\\mathsf{q}$ a priori independent and standard normal distributed, i.e. $\\mathsf{q} \\sim \\mathcal{N}(0, \\mathbb{I}_{6+2\\mathtt{ST}})$.\n",
"\n",
"A function to sample from the overall generative model given a latent input vector $\\mathsf{q}$ can be implemented using the `generate_z`, `generate_x_0` and `forward_func` functions and the JAX [`scan`](https://jax.readthedocs.io/en/latest/_autosummary/jax.lax.scan.html) operator (a differentiable loop / iterator construct) as follows"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [],
"source": [
"def generate_from_model(q, δ, dim_x, dim_z, dim_v, num_steps_per_obs):\n",
" \"\"\"Generate parameters and state + obs. sequences from model given q.\"\"\"\n",
" u, v_0, v_r = jnp.split(q, (dim_z, dim_z + dim_x))\n",
" z = generate_z(u)\n",
" x_0 = generate_x_0(z, v_0)\n",
" v_seq = jnp.reshape(v_r, (-1, dim_v))\n",
"\n",
" # Define integrator step function to scan:\n",
" # first argument is carried-forward state,\n",
" # second argument is input from scanned sequence.\n",
" def step_func(x, v):\n",
" x_n = forward_func(z, x, v, δ)\n",
" # Scan expects to return a tuple with the first element the carry-forward state\n",
" # and second element a slice of the output sequence (here also the state)\n",
" return x_n, x_n\n",
"\n",
" # Scan step_func over the noise sequence v_seq initialising carry-forward with x_init\n",
" _, x_seq = lax.scan(step_func, x_0, v_seq)\n",
" y_seq = obs_func(x_seq[num_steps_per_obs - 1 :: num_steps_per_obs])\n",
" return x_seq, y_seq, z, x_0\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Generate simulated observed data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"In order to allow us to illustrate performing inference with the model, we first generate simulated observed data from the model itself, with the aim of then inferring the posterior distribution on the 'unknown' latent state given the simulated observations. We use $\\mathtt{T} = 100$ observation times with interobservation interval $\\Delta = 0.5$ and $\\mathtt{S} = 25$ integrator steps per interobservation interval ($\\delta = 0.02$) giving us an overall latent dimension of $\\mathtt{Q} = 5006$ (we instead use $\\mathtt{T} = 20$ and $\\mathtt{S} = 10$ if running on Binder to reduce the CPU demand with then $\\mathtt{Q} = 406$)."
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"obs_interval = 0.5\n",
"if not ON_BINDER:\n",
" num_obs = 100\n",
" num_steps_per_obs = 25\n",
"else:\n",
" num_obs = 20\n",
" num_steps_per_obs = 10\n",
"num_steps = num_obs * num_steps_per_obs\n",
"dim_q = dim_z + dim_x + num_obs * num_steps_per_obs * dim_v\n",
"δ = obs_interval / num_steps_per_obs\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We seed a NumPy `RandomState` pseudo-random number generator object and use it to generate a latent input vector $\\mathsf{q}$ from its standard normal prior."
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [],
"source": [
"seed = 20250528\n",
"rng = onp.random.default_rng(seed)\n",
"q_ref = rng.standard_normal(size=dim_q)\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Using the previously defined `generate_from_model` function we now generate simulated state and observation sequences, parameters and the initial state from the model given the just generated latent input vector."
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"WARNING:2025-05-28 16:44:11,093:jax._src.xla_bridge:791: An NVIDIA GPU may be present on this machine, but a CUDA-enabled jaxlib is not installed. Falling back to cpu.\n"
]
}
],
"source": [
"x_seq_ref, y_seq_ref, z_ref, x_0_ref = generate_from_model(\n",
" q_ref, δ, dim_x, dim_z, dim_v, num_steps_per_obs\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We can visualise the simulated state and observation sequences using Matplotlib. Below the blue and orange lines show the time courses of respectively the $\\mathsf{x}_0$ and $\\mathsf{x}_1$ state components, with the blue crosses indicating the simulated discrete time observations of the $\\mathsf{x}_0$ component."
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"
"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"with plt.style.context(plot_style):\n",
" fig, axes = plt.subplots(2, sharex=True, figsize=(6, 4), dpi=150)\n",
" t_seq = (1 + onp.arange(num_steps)) * (num_obs) * obs_interval / num_steps\n",
" obs_indices = (1 + onp.arange(num_obs)) * num_steps_per_obs - 1\n",
" axes[0].plot(t_seq, x_seq_ref[:, 0], lw=0.5, color=\"C0\")\n",
" axes[1].plot(t_seq, x_seq_ref[:, 1], lw=0.5, color=\"C1\")\n",
" axes[0].plot(t_seq[obs_indices], y_seq_ref[:, 0], \"x\", ms=3, color=\"red\")\n",
" axes[0].set_ylabel(r\"$\\mathsf{x}_{0}$\")\n",
" axes[1].set_ylabel(r\"$\\mathsf{x}_{1}$\")\n",
" for ax in axes:\n",
" ax.set_xlim(0, num_obs * obs_interval)\n",
" _ = axes[1].set_xlabel(\"Time $\\\\tau$\")\n",
" fig.tight_layout()\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Manifold Markov chain Monte Carlo approximate inference using *Mici*\n",
"\n",
"[](https://matt-graham.github.io/mici/)\n",
"\n",
"To perform inference in the model given our simulated observed data, we use the manifold MCMC method implementations in the package [*Mici*](https://matt-graham.github.io/mici/). \n",
"\n",
"The key model-specific object required for inference in Mici is a *Hamiltonian system* instance. The Hamiltonian system encapsulates the various components of the Hamiltonian function for which the associated Hamiltonian dynamics are used as a proposal generating mechanism in a MCMC method. Mici includes various generic Hamiltonian system classes in the `mici.systems` module corresponding to common cases such as (unconstrained) systems with Euclidean and Riemannian metrics and constrained Hamiltonian systems with a constraint function with dense Jacobian. Here we instead use a custom system class defined in the `sde.mici_extensions` module which defines a constrained Hamiltonian system corresponding to a generative model for a diffusion as defined above (see Sections 3 and 4 in the paper). In particular our implementation exploits the sparsity induced in the Jacobian of the constraint function by artificially conditioning on the full state at a set of time points when sampling, as described in Section 5 in the paper. To construct an instance of this system class we pass in the variables defining the model dimensions defined earlier, the simulated observation sequence `y_seq_ref`, the generated `forward_func` implementing the strong-order 1.5 numerical integration scheme for the model, the `generate_x_0` and `generate_z` generator functions and `obs_func` observation function. This class expects the passed functions to be defined using JAX primitives such as via calls to functions in the `jax.numpy` module, so that it can use JAX's automatic differentiation primitives to automatically construct the required derivative functions."
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [],
"source": [
"num_obs_per_subseq = 5 # Number of obs in each fully conditioned subsequence\n",
"system = mici_extensions.ConditionedDiffusionConstrainedSystem(\n",
" obs_interval,\n",
" num_steps_per_obs,\n",
" num_obs_per_subseq,\n",
" y_seq_ref,\n",
" dim_z,\n",
" dim_x,\n",
" dim_v,\n",
" forward_func,\n",
" generate_x_0,\n",
" generate_z,\n",
" obs_func,\n",
" use_gaussian_splitting=True,\n",
")\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"As well as the Hamiltonian system we also need to define an associated (symplectic) integrator, to numerically simulate the associated Hamiltonian dynamics. Here we use the `mici.integrators.ConstrainedLeapfrogIntegrator` class, which corresponds to the constrained symplectic integrator described in Algorithm 1 in the paper (here we use the Gaussian specific Hamiltonian splitting described in Section 4.3.1 in the paper). We specify the tolerances on both the norm of the constraint equation `constraint_tol` and the successive change in the position `position_tol` for the Newton iteration used to solve the non-linear system of constraint equations, and also set a maximum number of iterations `max_iters`. The tolerances for the reversibility check is set to `2 * position_tol` (motivated by the intuition that each of the forward and backward retraction / projection steps are solved to a position tolerance of `position_tol`, so if the errors accumulate linearly the overall error in a reversible step should be less than `2 * position_tol`)."
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [],
"source": [
"max_iters = 50 # Maximum number of quasi-Newton iterations in retraction solver\n",
"constraint_tol = 1e-9 # Convergence tolerance in constraint (observation) space\n",
"position_tol = 1e-8 # Convergence tolerance in position (latent) space\n",
"integrator = mici.integrators.ConstrainedLeapfrogIntegrator(\n",
" system,\n",
" projection_solver=mici_extensions.jitted_solve_projection_onto_manifold_newton,\n",
" reverse_check_tol=2 * position_tol,\n",
" projection_solver_kwargs=dict(\n",
" constraint_tol=constraint_tol, position_tol=position_tol, max_iters=max_iters\n",
" ),\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The final key object required for inference in Mici, is a MCMC sampler class instance. Here we use a MCMC method which sequentially applies three Markov transition kernels leaving the (extended) target distribution invariant on each iteration. \n",
"\n",
"The first is a transition in which the momentum is independently resampled from its conditional distribution given the position (as described in Section 4.2 in the paper), as implemented by the `mici.transitions.IndependentMomentumTransition` class. We could instead for example use an instance of `mici.transitions.CorrelatedMomentumTransition` which would give to partial / correlated momentum resampling.\n",
"\n",
"The second transition is the main Hamiltonian-dynamics driven transition which simulates the Hamiltonian dynamics associated with the passed `system` object using the `integrator` object to generate proposed moves. Here we use `mici.transitions.MultinomialDynamicIntegrationTransition`, a dynamic integration time Hamiltonian Monte Carlo transition with multonimial sampling from the trajectory, analagous to the sampling algorithm used in the popular probabilistic programming framework [Stan](https://mc-stan.org/) and as described in Appendix A in the article [*A conceptual introduction to Hamiltonian Monte Carlo* (Betancourt, 2017)](https://arxiv.org/abs/1701.02434).\n",
"\n",
"The previous transition simulates the Hamiltonian dynamics for the *conditioned* diffusion system, i.e. full conditioning on a subset set of the states at the observation times. Therefore the third and final transition deterministically updates the set of observation time indices that are conditioned on in the Hamiltonian-dynamics integration based transition, here switching between two sets of observation time indices (partitions of the observation sequence) as descibed in Section 5 in the paper."
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {},
"outputs": [],
"source": [
"sampler = mici.samplers.MarkovChainMonteCarloMethod(\n",
" rng,\n",
" transitions={\n",
" \"momentum\": mici.transitions.IndependentMomentumTransition(system),\n",
" \"integration\": mici.transitions.MultinomialDynamicIntegrationTransition(\n",
" system, integrator\n",
" ),\n",
" \"switch_partition\": mici_extensions.SwitchPartitionTransition(system),\n",
" },\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"To generate a set of initial states on satisfying the observation constraints, we use a linear interpolation based scheme. A set of parameters $\\mathsf{z}$ and initial state $\\mathsf{x}_0$ are sampled from their prior distributions and a sequence of diffusion states at the observation time indices $\\tilde{\\mathsf{x}}_{1{:}\\mathtt{T}}$ sampled consistent with the observed sequence $\\mathsf{y}_{1{:}\\mathtt{T}}$ (i.e. such that $y_\\mathtt{t} = h_\\mathtt{t}(\\tilde{x}_\\mathtt{t}) ~~\\forall \\mathtt{t}\\in 1{:}\\mathtt{T}$). The sequence of noise vectors $\\mathsf{v}_{1{:}\\mathtt{ST}}$ which maps to a state sequence $\\mathsf{x}_{1{:}\\mathtt{ST}}$ which linear interpolates between the states in $\\tilde{\\mathsf{x}}_{1{:}\\mathtt{T}}$. This scheme requires that the forward function $f_\\delta$ is linear in the noise vector argument $\\mathsf{v}$ and that the Jacobian of $f_\\delta$ with respect to $\\mathsf{v}$ is full row-rank.\n",
"\n",
"\n",
"Due to the simple form of the observation function assumed here, to generate a diffusion state sequence $\\tilde{x}_{1{:}\\mathtt{T}}$ consistent with the observations $\\mathsf{y}_{1{:}\\mathtt{T}}$ we simply sample values for the $\\mathsf{x}_1$ components from $\\mathcal{N}(0, 0.5^2)$ and set the $\\mathsf{x}_0$ components values to the corresponding $y_{1{:}\\mathtt{T}}$ value. This is implemented in the function `generate_x_obs_seq_init` below."
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [],
"source": [
"def generate_x_obs_seq_init(rng):\n",
" return jnp.concatenate((y_seq_ref, rng.standard_normal(y_seq_ref.shape) * 0.5), -1)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We now generate a list of initial states, one for each of the chains to be run, using a helper function `find_initial_state_by_linear_interpolation` defined in the `sde.mici_extensions` module which implements the scheme described above."
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [],
"source": [
"num_chains = 2 # Number of independent Markov chains to run\n",
"init_states = [\n",
" mici_extensions.find_initial_state_by_linear_interpolation(\n",
" system, rng, generate_x_obs_seq_init\n",
" )\n",
" for _ in range(num_chains)\n",
"]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"As a final step before sampling the chains we define a function which outputs the variables to be traced (recorded) on each chain iteration."
]
},
{
"cell_type": "code",
"execution_count": 20,
"metadata": {},
"outputs": [],
"source": [
"def trace_func(state):\n",
" q = state.pos\n",
" u, v_0, v_seq = onp.split(q, (dim_z, dim_z + dim_x,))\n",
" v_seq = v_seq.reshape((-1, dim_v))\n",
" z = generate_z(u)\n",
" x_0 = generate_x_0(z, v_0)\n",
" return {\"x_0\": x_0, \"σ\": z[0], \"ϵ\": z[1], \"γ\": z[2], \"β\": z[3], \"v_seq\": v_seq}\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We now use the constructed `sampler` object to (sequentially) sample `num_chains` Markov chains for `num_warm_up_iter + n_main_iter` iterations. The first `n_warm_up_iter` iterations are an adaptive *warm up* stage used to tune the integrator step size (to give a target acceptance statistic of 0.9) and are not used when calculating estimates / statistics using the chain samples. We specify for four statistics to be monitored during sampling - the average acceptance statistic (`accept_stat`), proportion of integration transitions terminating due to non-convergence of the quasi-Newton iteration (`convergence_error`), the proportion of integration transitions terminating due to detection of a non-reversible step (`non_reversible_step`) and the number of integrator steps computed per transition (`n_step`).\n",
"\n",
"Due to the just-in-time compilation of the JAX model functions, the first couple of chain iterations will take longer as each of the model functions are compiled on their first calls (this happens for the first two rather than one iteration as the compiled model functions are specific to the partition / set of observation times conditioned on). During sampling, progress bars will be shown for each chain.\n",
"\n",
"Note as sampling the chains puts a high demand on the CPU we default to sampling only very short chains if running on Binder to avoid creating excessive CPU load on their servers (chains will also run much slower on Binder servers due to the restricted CPU availabity). We recommend running longer chains on your local machine; the default settings of 2 chains of 1000 samples took approximately 15 minutes to run on the laptop used for testing."
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
"