Skip to content

Commit

Permalink
fixed tutorial notebooks
Browse files Browse the repository at this point in the history
  • Loading branch information
Thomas Morris committed Oct 24, 2023
1 parent 648337a commit bbdf4ab
Show file tree
Hide file tree
Showing 3 changed files with 176 additions and 69 deletions.
144 changes: 114 additions & 30 deletions docs/source/tutorials/himmelblau.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,18 @@
"Let's use ``bloptools`` to minimize Himmelblau's function, which has four global minima:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "cf27fc9e-d11c-40f4-a200-98e7814f506b",
"metadata": {},
"outputs": [],
"source": [
"from bloptools.utils import prepare_re_env\n",
"\n",
"%run -i $prepare_re_env.__file__ --db-type=temp"
]
},
{
"cell_type": "code",
"execution_count": null,
Expand All @@ -29,16 +41,14 @@
"import numpy as np\n",
"import matplotlib as mpl\n",
"from matplotlib import pyplot as plt\n",
"from bloptools import test_functions\n",
"from bloptools.utils import functions\n",
"\n",
"x1 = x2 = np.linspace(-6, 6, 1024)\n",
"X1, X2 = np.meshgrid(x1, x2)\n",
"from bloptools.tasks import Task\n",
"\n",
"task = Task(key=\"himmelblau\", kind=\"min\")\n",
"F = test_functions.himmelblau(X1, X2)\n",
"F = functions.himmelblau(X1, X2)\n",
"\n",
"plt.pcolormesh(x1, x2, F, norm=mpl.colors.LogNorm())\n",
"plt.pcolormesh(x1, x2, F, norm=mpl.colors.LogNorm(vmin=1e-1, vmax=1e3), cmap=\"magma_r\")\n",
"plt.colorbar()\n",
"plt.xlabel(\"x1\")\n",
"plt.ylabel(\"x2\")"
Expand All @@ -59,11 +69,13 @@
"metadata": {},
"outputs": [],
"source": [
"from bloptools import devices\n",
"from bloptools.bayesian import DOF, BrownianMotion\n",
"\n",
"dofs = [\n",
" {\"device\": devices.DOF(name=\"x1\"), \"limits\": (-6, 6), \"kind\": \"active\"},\n",
" {\"device\": devices.DOF(name=\"x2\"), \"limits\": (-6, 6), \"kind\": \"active\"},\n",
" DOF(name=\"x1\", limits=(-6, 6)),\n",
" DOF(name=\"x2\", limits=(-6, 6)),\n",
" DOF(BrownianMotion(name=\"brownian1\"), read_only=True),\n",
" DOF(BrownianMotion(name=\"brownian2\"), read_only=True),\n",
"]"
]
},
Expand All @@ -82,9 +94,9 @@
"metadata": {},
"outputs": [],
"source": [
"tasks = [\n",
" {\"key\": \"himmelblau\", \"kind\": \"minimize\"},\n",
"]"
"from bloptools.bayesian import Objective\n",
"\n",
"objectives = [Objective(key=\"himmelblau\", minimize=True)]"
]
},
{
Expand All @@ -107,7 +119,7 @@
" products = db[uid].table()\n",
"\n",
" for index, entry in products.iterrows():\n",
" products.loc[index, \"himmelblau\"] = test_functions.himmelblau(entry.x1, entry.x2)\n",
" products.loc[index, \"himmelblau\"] = functions.himmelblau(entry.x1, entry.x2)\n",
"\n",
" return products"
]
Expand All @@ -130,54 +142,62 @@
},
"outputs": [],
"source": [
"from bloptools.utils import prepare_re_env\n",
"\n",
"%run -i $prepare_re_env.__file__ --db-type=temp\n",
"from bloptools.bayesian import Agent\n",
"\n",
"\n",
"agent = Agent(\n",
" dofs=dofs,\n",
" tasks=tasks,\n",
" objectives=objectives,\n",
" digestion=digestion,\n",
" db=db,\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "e964d5a5-2a4a-4403-8c06-4ad17b00cecf",
"metadata": {},
"outputs": [],
"source": [
"agent.test_inputs_grid().shape"
]
},
{
"cell_type": "markdown",
"id": "b7a608d6",
"id": "27685849",
"metadata": {},
"source": [
"To decide which points to sample, the agent needs an acquisition function. The available acquisition function are here:"
"Without any data, we can't make any inferences about what the function looks like, and so we can't use any non-trivial acquisition functions. Let's start by quasi-randomly sampling the parameter space, and plotting our model of the function:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "fb06739b",
"id": "996da937",
"metadata": {},
"outputs": [],
"source": [
"agent.acq_func_info"
"RE(agent.learn(\"quasi-random\", n=32))\n",
"agent.plot_objectives()"
]
},
{
"cell_type": "markdown",
"id": "27685849",
"id": "dc264346-10fb-4c88-9925-4bfcf0dd3b07",
"metadata": {},
"source": [
"Without any data, we can't make any inferences about what the function looks like, and so we can't use any non-trivial acquisition functions. Let's start by quasi-randomly sampling the parameter space, and plotting our model of the function:"
"To decide which points to sample, the agent needs an acquisition function. The available acquisition function are here:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "996da937",
"id": "fb06739b",
"metadata": {},
"outputs": [],
"source": [
"RE(agent.learn(\"quasi-random\", n=32))\n",
"agent.plot_tasks()"
"agent.acq_func_info"
]
},
{
Expand All @@ -196,9 +216,43 @@
"metadata": {},
"outputs": [],
"source": [
"agent.plot_acquisition(acq_funcs=[\"qei\", \"qpi\", \"ucb\"])"
"agent.plot_acquisition(acq_funcs=[\"qei\", \"pi\", \"qucb\"])"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9c0b42dc-2df3-4ba6-b02f-569dab48db80",
"metadata": {},
"outputs": [],
"source": [
"agent.dofs.limits"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "62b24d9b-740f-45a6-9617-47796c260273",
"metadata": {},
"outputs": [],
"source": [
"self = agent\n",
"import torch\n",
"\n",
"acq_func_lower_bounds = [dof.lower_limit if not dof.read_only else dof.readback for dof in self.dofs]\n",
"acq_func_upper_bounds = [dof.upper_limit if not dof.read_only else dof.readback for dof in self.dofs]\n",
"\n",
"torch.tensor(np.vstack([acq_func_lower_bounds, acq_func_upper_bounds]))"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "217158c0-aa65-409c-a7ab-63bb924723ad",
"metadata": {},
"outputs": [],
"source": []
},
{
"attachments": {},
"cell_type": "markdown",
Expand All @@ -208,6 +262,24 @@
"To decide where to go, the agent will find the inputs that maximize a given acquisition function:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "16ec3c97-211b-49df-9e45-fcdd61ae98eb",
"metadata": {},
"outputs": [],
"source": [
"agent.acquisition_function_bounds"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "a066da53-0cdc-429b-a588-ce22b4a599b5",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
Expand Down Expand Up @@ -238,9 +310,13 @@
},
"outputs": [],
"source": [
"X, _ = agent.ask(\"qei\", n=8)\n",
"X, _ = agent.ask(\"qei\", n=8, route=True)\n",
"agent.plot_acquisition(acq_funcs=[\"qei\"])\n",
"plt.plot(*X.T, lw=5e-1, c=\"r\", marker=\"x\")"
"plt.scatter(*X.T, marker=\"d\", facecolor=\"w\", edgecolor=\"k\")\n",
"plt.plot(\n",
" *X.T,\n",
" color=\"r\",\n",
")"
]
},
{
Expand Down Expand Up @@ -276,9 +352,17 @@
"metadata": {},
"outputs": [],
"source": [
"agent.plot_tasks()\n",
"print(agent.best_inputs)"
"agent.plot_objectives()\n",
"# print(agent.best_inputs)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "b6453b87-f864-40af-ba70-9a42960f54b9",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
Expand Down
26 changes: 13 additions & 13 deletions docs/source/tutorials/hyperparameters.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -21,12 +21,12 @@
"import numpy as np\n",
"import matplotlib as mpl\n",
"from matplotlib import pyplot as plt\n",
"from bloptools import test_functions\n",
"from bloptools.utils import functions\n",
"\n",
"x1 = x2 = np.linspace(-10, 10, 256)\n",
"X1, X2 = np.meshgrid(x1, x2)\n",
"\n",
"F = test_functions.booth(X1, X2)\n",
"F = functions.booth(X1, X2)\n",
"\n",
"plt.pcolormesh(x1, x2, F, norm=mpl.colors.LogNorm(), shading=\"auto\")\n",
"plt.colorbar()\n",
Expand Down Expand Up @@ -54,7 +54,7 @@
" products = db[uid].table()\n",
"\n",
" for index, entry in products.iterrows():\n",
" products.loc[index, \"booth\"] = test_functions.booth(entry.x1, entry.x2)\n",
" products.loc[index, \"booth\"] = functions.booth(entry.x1, entry.x2)\n",
"\n",
" return products"
]
Expand All @@ -72,28 +72,28 @@
"\n",
"%run -i $prepare_re_env.__file__ --db-type=temp\n",
"\n",
"from bloptools import devices\n",
"from bloptools.bayesian import Agent\n",
"from bloptools.bayesian import DOF, Objective, Agent\n",
"\n",
"dofs = [\n",
" {\"device\": devices.DOF(name=\"x1\"), \"limits\": (-5, 5), \"kind\": \"active\"},\n",
" {\"device\": devices.DOF(name=\"x2\"), \"limits\": (-5, 5), \"kind\": \"active\"},\n",
" DOF(name=\"x1\", limits=(-6, 6)),\n",
" DOF(name=\"x2\", limits=(-6, 6)),\n",
"]\n",
"\n",
"tasks = [\n",
" {\"key\": \"booth\", \"kind\": \"minimize\"},\n",
"objectives = [\n",
" Objective(key=\"booth\", minimize=True),\n",
"]\n",
"\n",
"\n",
"agent = Agent(\n",
" dofs=dofs,\n",
" tasks=tasks,\n",
" objectives=objectives,\n",
" digestion=digestion,\n",
" db=db,\n",
")\n",
"\n",
"RE(agent.learn(acq_func=\"qr\", n=16))\n",
"\n",
"agent.plot_tasks()"
"agent.plot_objectives()"
]
},
{
Expand Down Expand Up @@ -125,13 +125,13 @@
"outputs": [],
"source": [
"RE(agent.learn(\"qei\", n=4, iterations=4))\n",
"agent.plot_tasks()"
"agent.plot_objectives()"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3.11.4 64-bit",
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
Expand Down
Loading

0 comments on commit bbdf4ab

Please sign in to comment.