Skip to content

Commit

Permalink
Merge pull request #26 from MSDLLCpapers/release_docs
Browse files Browse the repository at this point in the history 
Add more Wiki pages and Tutorials
  • Loading branch information
@kstone40
kstone40 authored Aug 15, 2024
2 parents 06c4e1f + a744522 commit 8145b8f
Show file tree
Hide file tree
Showing 15 changed files with 18,111 additions and 3,779 deletions.
134 changes: 63 additions & 71 deletions demo/Constrained multi objective.ipynb → demo/Constrained multi-output min-max.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -7,34 +7,47 @@
"outputs": [],
"source": [
"import obsidian\n",
"obsidian.__version__"
"print(f'obsidian version: ' + obsidian.__version__)\n",
"\n",
"import pandas as pd\n",
"import plotly.express as px\n",
"import plotly.io as pio\n",
"pio.renderers.default = \"plotly_mimetype+notebook\""
]
},
{
"cell_type": "code",
"execution_count": null,
"cell_type": "markdown",
"metadata": {},
"outputs": [],
"source": [
"import pandas as pd\n",
"import plotly.express as px"
"## Introduction"
]
},
{
"cell_type": "code",
"execution_count": null,
"cell_type": "markdown",
"metadata": {},
"outputs": [],
"source": [
"from obsidian.parameters import ParamSpace, Param_Continuous\n",
"from obsidian.experiment import ExpDesigner"
"In this tutorial, we will see how to use _obsidian_ for multi-output optimization. To demonstrate the versatility of the approach, we will seek to maximize one response while minimizing the other.\n",
"\n",
"$$\\underset{X}{argmax} HV\\left(+f\\left(y_1\\right) -f\\left(y_2\\right)\\right)$$\n",
"\n",
"Furthermore, we will apply a linear constraint on the input variables; requiring that the $X_1 + X_2 \\leq 6 $."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Set up parameter space and initialize a design"
"## Set up parameter space and initialize a design"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from obsidian import Campaign, Target, ParamSpace, BayesianOptimizer\n",
"from obsidian.parameters import Param_Continuous"
]
},
{
Expand All @@ -49,8 +62,12 @@
" ]\n",
"\n",
"X_space = ParamSpace(params)\n",
"designer = ExpDesigner(X_space, seed=0)\n",
"X0 = designer.initialize(4, 'LHS')\n",
"target = [\n",
" Target('Response 1', aim='max'),\n",
" Target('Response 2', aim='min')\n",
"]\n",
"campaign = Campaign(X_space, target, seed=0)\n",
"X0 = campaign.designer.initialize(4, 'LHS')\n",
"\n",
"X0"
]
Expand All @@ -59,7 +76,7 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"### Collect results (e.g. from a simulation)"
"## Collect results (e.g. from a simulation)"
]
},
{
Expand All @@ -75,23 +92,15 @@
"y0 = simulator.simulate(X0)\n",
"Z0 = pd.concat([X0, y0], axis=1)\n",
"\n",
"Z0"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"Z0.plot(x='Response 1', y='Response 2', kind='scatter', figsize=(4,3))"
"campaign.add_data(Z0)\n",
"campaign.data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Create a campaign to track optimization"
"### Fit an optimizer and visualize results"
]
},
{
Expand All @@ -100,7 +109,7 @@
"metadata": {},
"outputs": [],
"source": [
"from obsidian.campaign import Campaign"
"campaign.fit()"
]
},
{
Expand All @@ -109,25 +118,23 @@
"metadata": {},
"outputs": [],
"source": [
"my_campaign = Campaign(X_space)\n",
"my_campaign.add_data(Z0)\n",
"my_campaign.data"
"from obsidian.plotting import surface_plot, optim_progress"
]
},
{
"cell_type": "markdown",
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"### Fit an optimizer"
"surface_plot(campaign.optimizer)"
]
},
{
"cell_type": "code",
"execution_count": null,
"cell_type": "markdown",
"metadata": {},
"outputs": [],
"source": [
"from obsidian.parameters import Target"
"## Optimize new experiment suggestions"
]
},
{
Expand All @@ -136,29 +143,16 @@
"metadata": {},
"outputs": [],
"source": [
"target = [\n",
" Target('Response 1', aim='max'),\n",
" Target('Response 2', aim='min')\n",
"]\n",
"\n",
"my_campaign.set_target(target)\n",
"my_campaign.fit()\n"
"from obsidian.constraints import InConstraint_Generic"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Make new experiment suggestions"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from obsidian.constraints import InConstraint_Generic"
"_Note:_ It is a good idea to balance a set of acquisition functions with those that prefer design-space exploration. This helps to ensure that the optimizer is not severely misled by deficiencies in the dataset, particularly for small data. It also helps to ascertain a global optimum.\n",
"\n",
"A simple choice is __Space Filling (SF)__ although __Negative Integrated Posterior Variance (NIPV)__ is available for single-output optimizations; and there are various other acquisiiton functions whose hyperparameters can be tuned to manage the \"explore-exploit\" balance."
]
},
{
Expand All @@ -167,13 +161,9 @@
"metadata": {},
"outputs": [],
"source": [
"# # X1 + X2 >= 2\n",
"# optim_kwargs = {'m_batch':2, 'acquisition':[{'qNEHVI':{'ref_point':[-350,0]}}], 'ineq_constraints':[InConstraint_Generic(X_space, indices=[0,1], coeff=[1,1], rhs=2)]}\n",
"\n",
"# X1 + X2 <= 6 aka -X1 - X2 >= -6\n",
"optim_kwargs = {'m_batch':2, 'acquisition':[{'qNEHVI':{'ref_point':[-350,0]}}], 'ineq_constraints':[InConstraint_Generic(X_space, indices=[0,1], coeff=[-1,-1], rhs=-6)]}\n",
"\n",
"X_suggest, eval_suggest = my_campaign.optimizer.suggest(**optim_kwargs)"
"X_suggest, eval_suggest = campaign.optimizer.suggest(acquisition = [{'NEHVI':{'ref_point':[-350, -20]}}, 'SF'],\n",
" # X1 + X2 <= 6, written as -X1 - X2 >= -6\n",
" ineq_constraints = [InConstraint_Generic(X_space, indices=[0,1], coeff=[-1,-1], rhs=-6)])"
]
},
{
Expand All @@ -189,7 +179,7 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"### Collect data at new suggestions"
"## Collect data at new suggestions"
]
},
{
Expand All @@ -200,15 +190,15 @@
"source": [
"y_iter1 = pd.DataFrame(simulator.simulate(X_suggest))\n",
"Z_iter1 = pd.concat([X_suggest, y_iter1, eval_suggest], axis=1)\n",
"my_campaign.add_data(Z_iter1)\n",
"my_campaign.data"
"campaign.add_data(Z_iter1)\n",
"campaign.data.tail()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Repeat as desired"
"## Repeat as desired"
]
},
{
Expand All @@ -218,11 +208,12 @@
"outputs": [],
"source": [
"for iter in range(5):\n",
" my_campaign.fit()\n",
" X_suggest, eval_suggest = my_campaign.optimizer.suggest(**optim_kwargs)\n",
" campaign.fit()\n",
" X_suggest, eval_suggest = campaign.optimizer.suggest(acquisition = [{'NEHVI':{'ref_point':[-350, -20]}}, 'SF'],\n",
" ineq_constraints = [InConstraint_Generic(X_space, indices=[0,1], coeff=[-1,-1], rhs=-6)])\n",
" y_iter = pd.DataFrame(simulator.simulate(X_suggest))\n",
" Z_iter = pd.concat([X_suggest, y_iter, eval_suggest], axis=1)\n",
" my_campaign.add_data(Z_iter)"
" campaign.add_data(Z_iter)"
]
},
{
Expand All @@ -231,8 +222,7 @@
"metadata": {},
"outputs": [],
"source": [
"fig = px.scatter(my_campaign.data, x='Response 1', y='Response 2', color='Iteration')\n",
"fig.update_layout(height=300, width=400, template='ggplot2')"
"optim_progress(campaign)"
]
},
{
Expand All @@ -241,15 +231,17 @@
"metadata": {},
"outputs": [],
"source": [
"my_campaign.data"
"surface_plot(campaign.optimizer, response_id = 0)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
"source": [
"surface_plot(campaign.optimizer, response_id = 1)"
]
}
],
"metadata": {
Expand Down
Loading

0 comments on commit 8145b8f

Please sign in to comment.