initial commit

.gitignore

@@ -0,0 +1,13 @@
+*.pyc
+*.log
+*.out
+*.blg
+*.aux
+*.bib
+#*#
+*.DS_Store
+*.pdf
+*.py~
+
+
+

README.md

@@ -1,4 +1,21 @@
-spearmint
-=========
+spearmint: Bayesian optimization codebase
+=========================================
 
-Spearmint Bayesian optimization codebase for eventual release
+
+STEP 1: Installation<br>
+1) Install python, numpy, scipy. For academic users, the anaconda distribution is great. Use numpy 1.8 or higher. 
+2) Download/clone the spearmint code<br>
+3) Install the spearmint package using pip: "pip install -e \</path/to/spearmint/root\>" (the -e means changes will be reflected automatically)<br>
+4) Download and install MongoDB: https://www.mongodb.org/<br>
+5) Install the pymongo package using e.g., pip or anaconda
+
+STEP 2: Setting up your experiment<br>
+1) Create a callable objective function. See ../examples/branin/branin.py as an example<br>
+2) Create a config file. There are 3 example config files in the ../examples directory.
+
+STEP 3: Running spearmint<br>
+1) Start up a MongoDB daemon instance: mongod --fork --logpath \<path/to/logfile\> --dbpath \<path/to/dbfolder\><br>
+2) Run spearmint: "python main.py \</path/to/experiment/directory\>"
+
+STEP 4: Looking at your results
+Spearmint will output results to standard out / standard err. You can also load the results from the database and manipulate them directly. 

examples/constrained/branin_con.py

@@ -0,0 +1,35 @@
+import math
+import numpy as np
+
+def evaluate(job_id, params):
+
+    x = params['X']
+    y = params['Y']
+
+    print 'Evaluating at (%f, %f)' % (x, y)
+
+    if x < 0 or x > 5.0 or y > 5.0:
+        return np.nan
+    # Feasible region: x in [0,5] and y in [0,5]
+
+    obj = float(np.square(y - (5.1/(4*np.square(math.pi)))*np.square(x) + (5/math.pi)*x- 6) + 10*(1-(1./(8*math.pi)))*np.cos(x) + 10)
+
+    con1 = float(y-x)   # y >= x 
+
+    con2 = float(10.0-y)  # y <= 10
+
+    return {
+        "branin"       : obj, 
+        "y_at_least_x" : con1, 
+        "y_at_most_10" : con2
+    }
+
+    # True minimum is at 2.945, 2.945, with a value of 0.8447
+
+def main(job_id, params):
+    try:
+        return evaluate(job_id, params)
+    except Exception as ex:
+        print ex
+        print 'An error occurred in branin_con.py'
+        return np.nan

examples/constrained/config.json

@@ -0,0 +1,35 @@
+{
+    "language"        : "PYTHON",
+    "main-file"       : "branin_con",
+    "experiment-name" : "branin-constrained-example",
+    "variables": {
+        "X": {
+            "type": "FLOAT",
+            "size": 1,
+            "min":  -5,
+            "max":  10
+        },
+        "Y": {
+            "type": "FLOAT",
+            "size": 1,
+            "min":  0,
+            "max":  15
+        }
+    },
+    "tasks": {
+      "branin" : {
+          "type"        : "OBJECTIVE",
+          "likelihood"  : "NOISELESS"
+      },
+      "y_at_least_x" : {
+          "type"        : "CONSTRAINT",
+          "likelihood"  : "NOISELESS"
+      },
+      "y_at_most_10" : {
+          "type"        : "CONSTRAINT",
+          "likelihood"  : "NOISELESS"
+      }
+    },
+    "polling-time" : 1
+}
+

examples/distributed/branin.py

@@ -0,0 +1,24 @@
+import numpy as np
+import sys
+import math
+import time
+
+def branin(x, y):
+
+  result = np.square(y - (5.1/(4*np.square(math.pi)))*np.square(x) + 
+       (5/math.pi)*x - 6) + 10*(1-(1./(8*math.pi)))*np.cos(x) + 10
+
+  result = float(result)
+
+  #if np.random.rand > 0.75:
+  #  raise Exception('Blah!')
+
+  print 'Result = %f' % result
+  time.sleep(np.random.randint(30))
+  return {'branin' : result}
+
+# Write a function like this called 'main'
+def main(job_id, params):
+  print 'Anything printed here will end up in the output directory for job #%d' % job_id
+  print params
+  return branin(params['x'], params['y'])

examples/distributed/config.json

@@ -0,0 +1,40 @@
+{
+    "language"        : "PYTHON",
+    "experiment-name" : "branin-distributed-example",
+    "polling-time"    : 1,
+    "resources" : {
+        "my-machine" : {
+            "scheduler"         : "local",
+            "max-concurrent"    : 1,
+            "max-finished-jobs" : 100
+        },
+        "cluster" : {
+            "scheduler"         : "SLURM",
+            "max-concurrent"    : 3,
+            "max-finished-jobs" : 1000
+        }
+    },
+    "tasks": {
+        "branin" : {
+            "type"       : "OBJECTIVE",
+            "likelihood" : "NOISELESS",
+            "main-file"  : "branin",
+            "resources"  : ["my-machine", "cluster"]
+        }
+    },
+    "variables": {
+        "x" : {
+            "type" : "FLOAT",
+            "size" : 1,
+            "min"  : -5,
+            "max"  : 10
+        },
+        "y" : {
+            "type" : "FLOAT",
+            "size" : 1,
+            "min"  : 0,
+            "max"  : 15
+        }
+    }
+}
+

examples/simple/branin.py

@@ -0,0 +1,21 @@
+import numpy as np
+import sys
+import math
+import time
+
+def branin(x, y):
+
+  result = np.square(y - (5.1/(4*np.square(math.pi)))*np.square(x) + 
+       (5/math.pi)*x - 6) + 10*(1-(1./(8*math.pi)))*np.cos(x) + 10
+
+  result = float(result)
+
+  print 'Result = %f' % result
+  #time.sleep(np.random.randint(60))
+  return result
+
+# Write a function like this called 'main'
+def main(job_id, params):
+  print 'Anything printed here will end up in the output directory for job #%d' % job_id
+  print params
+  return branin(params['x'], params['y'])

examples/simple/config.json

@@ -0,0 +1,20 @@
+{
+    "language"        : "PYTHON",
+    "main-file"       : "branin.py",
+    "experiment-name" : "simple-braninhoo-example",
+    "variables" : {
+        "x" : {
+            "type" : "FLOAT",
+            "size" : 1,
+            "min"  : -5,
+            "max"  : 10
+        },
+        "y" : {
+            "type" : "FLOAT",
+            "size" : 1,
+            "min"  : 0,
+            "max"  : 15
+        }
+    }
+}
+