{ "cells": [ { "cell_type": "markdown", "id": "915e3035-fc65-4020-a0ac-48657d952f65", "metadata": {}, "source": [ "## Tutorial Part 3: Annealed Sequential-Monte-Carlo Sampler" ] }, { "cell_type": "markdown", "id": "5d466658-6b2c-4ee0-9f84-0828621df79b", "metadata": { "tags": [] }, "source": [ "![](figures/smcs_nvi.png)\n", "\n", "### Introduction\n", "\n", "In this tutorial we using inference combinators to implement a Sequential Monte Carlo Sampler [(Del Moral et al., 2006)](https://rss.onlinelibrary.wiley.com/doi/abs/10.1111/j.1467-9868.2006.00553.x) that generates samples along a geometric annealing path \n", "\n", "\\begin{align}\n", "\\gamma_k(x; \\beta_k) := q_1(x)^{(1-\\beta_k)}\\gamma(x)^{\\beta_k}\n", ",&&\n", "\\beta_1 = 0\n", ",&&\n", "\\beta_k \\in [0, 1]\n", ",&&\n", "\\beta_K = 1,\n", "\\end{align}\n", "\n", "as outlined in [Zimmermann et al.](https://proceedings.neurips.cc/paper/2021/file/ab49b208848abe14418090d95df0d590-Paper.pdf). \n", "At each step $k$ we are given approximate samples $x_{k-1}$ from the last target $\\gamma_{k-1}$ and use a variational proposals $q_k(\\cdot \\mid x_{k-1})$ to generate proposals $x_k$ for the next target density $\\gamma_k$. We use these proposals to compute an importance weights\n", "\n", "\\begin{align}\n", "w_1 = 1\n", ",&& \n", "w_k = \\frac{r_{k-1}(x_{k-1} \\mid x_k)\\gamma_k(x_k)}{\\gamma_{k-1}(x_{k-1})q_k(x_k \\mid x_{k-1})}\n", "&& \\text{for}\\ 1 < k\\leq K\n", ",\n", "\\end{align}\n", "\n", "which we can use to resample our particles in order to get approximate samples for the next target distribution $\\gamma_k$. \n", "\n", "While this sampling strategy is always valid, the quality of the sampler crucially depends on the variance of the importance weight above. The variance of the importance weights $w_k$ is minimized when \n", "\n", "\\begin{align}\n", "\\check \\gamma_k(x_{k-1}, x_k)\n", ":= r_{k-1}(x_{k-1} \\mid x_k)\\gamma_k(x_k) && \\text{and}\\ && \\hat \\gamma_k(x_{k-1}, x_k)\n", ":= \\gamma_{k-1}(x_{k-1})q_k(x_k \\mid x_{k-1})\n", "\\end{align}\n", "\n", "are equal. To make $\\hat \\gamma_k$ as similar as possible to $\\check\\gamma_k$, we model the variational proposals $q_k$ and reverse kernels $r_k$ as variational distribution with parameters $\\phi$ and $\\theta$ respectively and minimize a KL-divergence\n", "\n", "\\begin{align}\n", "\\mathcal{D}_{\\mathrm{KL}}(\\hat\\gamma_{k} \\mid \\check \\gamma_{k})\n", "=\n", "\\mathbb{E}_{\\hat\\gamma_{k}(x_{k-1}, x_{k}; \\phi, \\beta_{k-1})}\n", "\\left[\n", "\\log \\frac{\\hat \\gamma_k(x_{k-1}, x_k; \\phi, \\beta_{k-1})}{\\check\\gamma_k(x_{k-1}, x_k; \\theta, \\beta_{k})}\n", "\\right]\n", "\\end{align}\n", "\n", "at each step. We additionally learn the parameters of the schedule of the annealing path $\\beta_k$ (for $1" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "import matplotlib.pyplot as plt\n", "import numpy as np\n", "import jax.numpy as jnp\n", "import flax\n", "import flax.linen as nn\n", "import numpyro.distributions as dist\n", "import numpyro\n", "\n", "numpyro.set_platform(\"cpu\")\n", "\n", "\n", "def ring_gmm_log_density(x, M):\n", " angles = 2 * jnp.arange(1, M + 1) * jnp.pi / M\n", " mu = 10 * jnp.stack([jnp.sin(angles), jnp.cos(angles)], -1)\n", " sigma = jnp.sqrt(0.5)\n", " return nn.logsumexp(\n", " dist.Normal(mu, sigma).log_prob(x[..., None, :]).sum(-1), -1\n", " )\n", "\n", "\n", "def proposal_log_density(x):\n", " return dist.Normal(0, 5).log_prob(x).sum(-1)\n", "\n", "\n", "xrange = np.linspace(-12, 12, 100)\n", "m_xy = np.dstack(np.meshgrid(xrange, xrange))\n", "m_target = np.exp(ring_gmm_log_density(m_xy, M=8))\n", "m_proposal = np.exp(proposal_log_density(m_xy))\n", "\n", "fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2)\n", "ax1.set_title(\"Proposal Density\")\n", "ax1.imshow(m_proposal)\n", "xax1, yax1 = ax1.axes.get_xaxis(), ax1.axes.get_yaxis()\n", "xax1.set_visible(False)\n", "yax1.set_visible(False)\n", "ax2.set_title(\"Target Density\")\n", "ax2.imshow(m_target)\n", "xax2, yax2 = ax2.axes.get_xaxis(), ax2.axes.get_yaxis()\n", "xax2.set_visible(False)\n", "yax2.set_visible(False)" ] }, { "cell_type": "markdown", "id": "b2c37eb8-a6ac-4a60-9256-cb031f188c16", "metadata": {}, "source": [ "### Defining a Sequence of annealed intermediate densities\n", "\n", "Given the proposal and target densities defined above, we can define our intermediate annealed densities using flax. Note that we are defining the tunable parameters `beta_raw` in log space and normalize them appropriately to be in the interval $[0, 1]$." ] }, { "cell_type": "code", "execution_count": 3, "id": "85cd547d-ba52-4635-978f-d7fd710fce63", "metadata": { "tags": [] }, "outputs": [], "source": [ "class AnnealedDensity(nn.Module):\n", " M = 8\n", "\n", " @nn.compact\n", " def __call__(self, x, index=0):\n", " beta_raw = self.param(\"beta_raw\", lambda _: -jnp.ones(self.M - 2))\n", " beta = nn.sigmoid(\n", " beta_raw[0] + jnp.pad(jnp.cumsum(nn.softplus(beta_raw[1:])), (1, 0))\n", " )\n", " beta = jnp.pad(beta, (1, 1), constant_values=(0, 1))\n", " beta_k = beta[index]\n", "\n", " target_density = ring_gmm_log_density(x, self.M)\n", " init_proposal = proposal_log_density(x)\n", " return beta_k * target_density + (1 - beta_k) * init_proposal" ] }, { "cell_type": "markdown", "id": "e965b054-9eea-4881-ae32-8a8431835d7b", "metadata": {}, "source": [ "### Network\n", "\n", "We now define our variational kernels, which each consist of a two-layer MLP with ReLU activations. We additionally define a helper class `VariationalKernelList`, which gives us convenient access to the individual variational kernels (automatically selecting the correct set of parameters), by providing the corresponding index. Lastly we define a `Network` class which wraps around the annealed target, forward kernels, and reverse kernels. This is convenient as it lets us pass around a single object which gives us access to the network of our individual model components." ] }, { "cell_type": "code", "execution_count": 4, "id": "8e36102f-6061-4776-b85f-7165e731ecd3", "metadata": { "tags": [] }, "outputs": [], "source": [ "class VariationalKernelNetwork(nn.Module):\n", "\n", " @nn.compact\n", " def __call__(self, x):\n", " h = nn.Dense(50)(x)\n", " h = nn.relu(h)\n", " loc = nn.Dense(2, kernel_init=nn.initializers.zeros)(h) + x\n", " scale_raw = nn.Dense(2, kernel_init=nn.initializers.zeros)(h)\n", " return loc, nn.softplus(scale_raw)\n", "\n", "\n", "class VariationalKernelNetworks(nn.Module):\n", " M = 8\n", "\n", " @nn.compact\n", " def __call__(self, x, index=0):\n", " if self.is_mutable_collection('params'):\n", " vmap_net = nn.vmap(\n", " VariationalKernelNetwork,\n", " variable_axes={'params': 0},\n", " split_rngs={'params': True},\n", " )\n", " out = vmap_net(name='kernel')(\n", " jnp.broadcast_to(x, (self.M - 1,) + x.shape)\n", " )\n", " return jax.tree.map(lambda x: x[index], out)\n", " params = self.scope.get_variable('params', 'kernel')\n", " params_i = jax.tree.map(lambda x: x[index], params)\n", " return VariationalKernelNetwork(name='kernel').apply(\n", " flax.core.freeze({'params': params_i}), x\n", " )\n", "\n", "\n", "class Networks(nn.Module):\n", "\n", " def setup(self):\n", " self.forward_kernel_params = VariationalKernelNetworks()\n", " self.reverse_kernel_params = VariationalKernelNetworks()\n", " self.anneal_density = AnnealedDensity()\n", "\n", " def __call__(self, x):\n", " self.reverse_kernel_params(x)\n", " self.anneal_density(x)\n", " return self.forward_kernel_params(x)" ] }, { "cell_type": "markdown", "id": "000158d9-a7ed-44f1-805c-03b5e4d22b52", "metadata": {}, "source": [ "### Defining the model components\n", "\n", "With all of our network definitions in place we can now define our model components as probabilistic programs in numpyro." ] }, { "cell_type": "code", "execution_count": 5, "id": "cd7857c9-2f05-4681-92e5-a880dc73bcc2", "metadata": { "tags": [] }, "outputs": [], "source": [ "def anneal_target(network, k=0):\n", " x = numpyro.sample(\"x\", dist.Normal(0, 5).expand([2]).mask(False).to_event())\n", " # numpyro.factor(\"anneal_density\", network.anneal_density(x, index=k))\n", " numpyro.sample(\n", " \"anneal_density\", dist.Unit(network.anneal_density(x, index=k))\n", " )\n", " return ({\"x\": x},)\n", "\n", "\n", "def anneal_forward(network, inputs, k=0):\n", " mu, sigma = network.forward_kernel_params(inputs[\"x\"], index=k)\n", " return numpyro.sample(\"x\", dist.Normal(mu, sigma).to_event(1))\n", "\n", "\n", "def anneal_reverse(network, inputs, k=0):\n", " mu, sigma = network.reverse_kernel_params(inputs[\"x\"], index=k)\n", " return numpyro.sample(\"x\", dist.Normal(mu, sigma).to_event(1))" ] }, { "cell_type": "markdown", "id": "7a7c1c6e-89b7-4229-bfdf-0d0dd3c9b242", "metadata": {}, "source": [ "### Using predefined inference algorithms in coix\n", "\n", "Coix already implements a selection of inference algorithms including Nested Variational Inference (NVI). All we need to do is to instantiate our model components and pass it to the method that composes the inference program for us. " ] }, { "cell_type": "code", "execution_count": 6, "id": "3f5c1c7c-b67c-459f-b6b8-85ee734d0730", "metadata": { "tags": [] }, "outputs": [], "source": [ "from functools import partial\n", "import jax\n", "from jax import random\n", "import coix\n", "\n", "coix.set_backend(\"coix.numpyro\")\n", "\n", "\n", "def make_anneal(params, unroll=False, num_particles=10, num_targets=8):\n", " network = coix.util.BindModule(Networks(), params)\n", " # Add particle dimension and construct a program.\n", " make_particle_plate = lambda: numpyro.plate(\"particle\", num_particles, dim=-1)\n", " targets = lambda k: make_particle_plate()(\n", " partial(anneal_target, network, k=k)\n", " )\n", " forwards = lambda k: make_particle_plate()(\n", " partial(anneal_forward, network, k=k)\n", " )\n", " reverses = lambda k: make_particle_plate()(\n", " partial(anneal_reverse, network, k=k)\n", " )\n", " if unroll: # to unroll the algorithm, we provide a list of programs\n", " targets = [targets(k) for k in range(num_targets)]\n", " forwards = [forwards(k) for k in range(num_targets - 1)]\n", " reverses = [reverses(k) for k in range(num_targets - 1)]\n", " program = coix.algo.nvi_rkl(\n", " targets, forwards, reverses, num_targets=num_targets\n", " )\n", " return program" ] }, { "cell_type": "markdown", "id": "acecd0e3-fa29-45db-bd6e-15dac146e714", "metadata": { "tags": [] }, "source": [ "### Evaluating the untrained model\n", "\n", "As mentioned before, while our sampler might not be very efficient before we train it, it is still valid. So let's see how our sampler performs pre-training first." ] }, { "cell_type": "code", "execution_count": 7, "id": "83372a7c-1803-4373-9943-1fec013defcb", "metadata": { "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'ess': 16678.6855, 'log_Z': 2.0389, 'log_density': -3.656, 'loss': 12.3431}\n" ] }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from matplotlib.animation import FuncAnimation\n", "from matplotlib.patches import Ellipse, Rectangle\n", "\n", "\n", "def eval_program(seed, params, num_particles):\n", " with numpyro.handlers.seed(rng_seed=seed):\n", " p = make_anneal(params, unroll=True, num_particles=num_particles)\n", " out, trace, metrics = coix.traced_evaluate(p)()\n", " return out, trace, metrics\n", "\n", "\n", "anneal_net = Networks()\n", "init_params = anneal_net.init(random.PRNGKey(0), jnp.zeros(2))\n", "_, trace, metrics = eval_program(\n", " random.PRNGKey(1), init_params, num_particles=100000\n", ")\n", "\n", "metrics.pop(\"log_weight\")\n", "anneal_metrics = jax.tree.map(\n", " lambda x: round(float(jnp.mean(x)), 4), metrics\n", ")\n", "print(anneal_metrics)\n", "\n", "fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2)\n", "x = trace[\"x\"][\"value\"].reshape((-1, 2))\n", "H, xedges, yedges = np.histogram2d(\n", " x[:, 0], x[:, 1], range=[[-12, 12], [-12, 12]], bins=100\n", ")\n", "ax1.set_title(\"Untrained Proposal Density\")\n", "ax1.imshow(H.T)\n", "xax1, yax1 = ax1.axes.get_xaxis(), ax1.axes.get_yaxis()\n", "xax1.set_visible(False)\n", "yax1.set_visible(False)\n", "ax2.set_title(\"Target Density\")\n", "ax2.imshow(m_target)\n", "xax2, yax2 = ax2.axes.get_xaxis(), ax2.axes.get_yaxis()\n", "xax2.set_visible(False)\n", "yax2.set_visible(False)" ] }, { "cell_type": "markdown", "id": "6c4a6e2c-c5f7-434c-9e56-2b236f366245", "metadata": {}, "source": [ "While visually the samples loosely approximate the target density we can clearly see that the modes are too wide and not tightly enough peaked. This is also reflected in the statistics that we can access in the metics dictionary returned by the evaluation effect handler. We can see that the ess is around $300$ when taking $1000$ samples." ] }, { "cell_type": "markdown", "id": "5e214589-58e9-45f5-8057-042cc2f9bb89", "metadata": {}, "source": [ "### Training\n", "\n", "So let's see if we can do better by training our model with NVI. All we need to do is to repeatedly run our inference program, differentiate the resulting loss, and take a gradient step. `coix.util.train` provides a convenient wrapper for such a training loop and optionally compiles the training procedure." ] }, { "cell_type": "code", "execution_count": 8, "id": "5ef69930-2758-4464-a04a-349ca52afe47", "metadata": { "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Compiling the first train step...\n", "Time to compile a train step: 6.4172139167785645\n", "=====\n", "Step 2500 | ess 25.2412 | log_Z 1.2350 | log_density -3.7176 | loss 0.9471 | squared_grad_norm 8029.2070\n", "Step 5000 | ess 31.7915 | log_Z 2.5552 | log_density -2.5793 | loss 0.6548 | squared_grad_norm 3195.3689\n", "Step 7500 | ess 31.3730 | log_Z 1.9925 | log_density -2.5403 | loss 0.8958 | squared_grad_norm 1125.2384\n", "Step 10000 | ess 31.2693 | log_Z 1.4725 | log_density -3.0077 | loss 0.8856 | squared_grad_norm 252.5485\n", "Step 12500 | ess 32.1133 | log_Z 2.1890 | log_density -2.0786 | loss 1.3447 | squared_grad_norm 435.7955\n", "Step 15000 | ess 33.2006 | log_Z 1.8315 | log_density -2.8067 | loss 0.7987 | squared_grad_norm 306.8722\n", "Step 17500 | ess 34.6039 | log_Z 2.0502 | log_density -2.1592 | loss 1.7737 | squared_grad_norm 120.0838\n", "Step 20000 | ess 34.3501 | log_Z 2.1727 | log_density -2.1159 | loss 1.5451 | squared_grad_norm 58.9659\n", "Step 22500 | ess 34.5807 | log_Z 2.2028 | log_density -2.1809 | loss 1.5018 | squared_grad_norm 90.4855\n", "Step 25000 | ess 35.5768 | log_Z 2.3687 | log_density -2.0594 | loss 1.4789 | squared_grad_norm 87.8662\n", "Step 27500 | ess 34.8928 | log_Z 2.0132 | log_density -2.9821 | loss 0.7715 | squared_grad_norm 112.2418\n", "Step 30000 | ess 34.8616 | log_Z 1.6396 | log_density -2.2300 | loss 1.7796 | squared_grad_norm 53.1720\n", "Step 32500 | ess 35.5625 | log_Z 1.9240 | log_density -2.1137 | loss 0.8976 | squared_grad_norm 66.0242\n", "Step 35000 | ess 35.5778 | log_Z 2.0557 | log_density -2.2012 | loss 2.4074 | squared_grad_norm 68.5943\n", "Step 37500 | ess 35.5934 | log_Z 2.1605 | log_density -2.0528 | loss 1.3179 | squared_grad_norm 50.1874\n", "Step 40000 | ess 35.2403 | log_Z 2.2516 | log_density -2.2726 | loss 1.4254 | squared_grad_norm 53.3538\n", "Step 42500 | ess 35.2561 | log_Z 2.3036 | log_density -2.1802 | loss 1.9746 | squared_grad_norm 85.2577\n", "Step 45000 | ess 35.8931 | log_Z 2.1175 | log_density -1.8413 | loss 1.7932 | squared_grad_norm 71.1966\n", "Step 47500 | ess 35.5040 | log_Z 2.0723 | log_density -2.4489 | loss 1.2341 | squared_grad_norm 33.0538\n", "Step 50000 | ess 35.1216 | log_Z 2.0225 | log_density -2.3940 | loss 1.6225 | squared_grad_norm 82.7994\n" ] } ], "source": [ "import optax\n", "\n", "\n", "def loss_fn(params, key, num_particles, unroll=False):\n", " # Run the program and get metrics.\n", " program = make_anneal(params, num_particles=num_particles, unroll=unroll)\n", " with numpyro.handlers.seed(rng_seed=key):\n", " _, _, metrics = coix.traced_evaluate(program)()\n", " return metrics[\"loss\"], metrics\n", "\n", "\n", "optimizer = optax.adam(1e-3)\n", "num_steps = 50000\n", "num_particles = 36\n", "unroll = True\n", "\n", "trained_params, metrics = coix.util.train(\n", " partial(loss_fn, num_particles=num_particles, unroll=unroll),\n", " init_params,\n", " optimizer,\n", " num_steps,\n", " jit_compile=True,\n", ")" ] }, { "cell_type": "markdown", "id": "9fcfdd89-9b63-4bbc-b124-572336a5c6cb", "metadata": {}, "source": [ "While the training loss with only 36 particles is not very informative to assess convergence, we can see that the ess goes up to about $34$ which is a good indicator that we learned a good proposal." ] }, { "cell_type": "markdown", "id": "a603dc86-bb81-4038-b668-1145be14e0e9", "metadata": {}, "source": [ "### Evaluate trained model\n", "We already implemented an evaluation wrapped, so let's evaluate the model again but this time with the optimized parameters." ] }, { "cell_type": "code", "execution_count": 9, "id": "2723e40f-9708-4c83-a2dc-f3dd15719dd0", "metadata": { "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'ess': 97606.4062, 'log_Z': 2.0777, 'log_density': -2.1979, 'log_weight': 2.0639, 'loss': 1.5816}\n" ] }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "_, trace, metrics = eval_program(\n", " random.PRNGKey(1), trained_params, num_particles=100000\n", ")\n", "\n", "anneal_metrics = jax.tree.map(\n", " lambda x: round(float(jnp.mean(x)), 4), metrics\n", ")\n", "print(anneal_metrics)\n", "\n", "fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2)\n", "x = trace[\"x\"][\"value\"].reshape((-1, 2))\n", "m_proposal, _, _ = np.histogram2d(\n", " x[:, 0], x[:, 1], range=[[-12, 12], [-12, 12]], bins=100\n", ")\n", "ax1.set_title(\"Trained Proposal Density\")\n", "ax1.imshow(m_proposal.T)\n", "xax1, yax1 = ax1.axes.get_xaxis(), ax1.axes.get_yaxis()\n", "xax1.set_visible(False)\n", "yax1.set_visible(False)\n", "ax2.set_title(\"Target Density\")\n", "ax2.imshow(m_target)\n", "xax2, yax2 = ax2.axes.get_xaxis(), ax2.axes.get_yaxis()\n", "xax2.set_visible(False)\n", "yax2.set_visible(False)" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.11.4" } }, "nbformat": 4, "nbformat_minor": 5 }