Add Likelihood Mode

README.rst

@@ -1,5 +1,8 @@
-.. image:: https://travis-ci.org/gerbaudo/fbu.png
-   :target: https://travis-ci.org/gerbaudo/fbu
+.. image:: https://travis-ci.com/pyFBU/fbu.png
+   :target: https://travis-ci.com/pyFBU/fbu
+
+.. image:: https://img.shields.io/pypi/v/fbu.svg
+   :target: https://pypi.python.org/pypi/fbu
 
 =====
 PyFBU
@@ -31,6 +34,20 @@ PyMC 3 can be installed using conda
 
    conda install -c conda-forge pymc3
 
+The following packages also need to be installed
+
+::
+
+    conda install mkl
+    conda install numpy
+    conda install mkl-service
+
+And this export needs to be added to your bashrc or similar to avoid warings
+
+::
+
+    export MKL_THREADING_LAYER=GNU
+
 or pip
 
 ::
@@ -45,15 +62,17 @@ The latest stable version of PyFBU can be installed using pip.
     pip install fbu
 
 Alternatively one can check out the development version of the code from the 
-`GitHub <https://github.com/gerbaudo/fbu>`_ repository:
+`GitHub <https://github.com/pyFBU/fbu>`_ repository:
 
 ::
 
-	git clone https://github.com/gerbaudo/fbu.git
+	git clone https://github.com/pyFBU/fbu.git
 
 
 Usage
 -----
 
-A `simple tutorial <https://github.com/gerbaudo/fbu/blob/master/tutorial.ipynb>`_ to help you get started.
+A `simple tutorial <tutorial.ipynb>`_ to help you get started.
+
+
 

fbu/PyFBU.py

@@ -1,6 +1,7 @@
 import pymc3 as mc
 from numpy import random, dot, array, inf
 import theano
+import copy
 
 class PyFBU(object):
     """A class to perform a MCMC sampling.
@@ -14,7 +15,7 @@ class PyFBU(object):
     def __init__(self,data=[],response=[],background={},
                  backgroundsyst={},objsyst={'signal':{},'background':{}},
                  lower=[],upper=[],regularization=None,
-                 rndseed=-1,verbose=False,name='',monitoring=False):
+                 rndseed=-1,verbose=False,name='',monitoring=False, mode=False):
         #                                     [MCMC parameters]
         self.nTune = 1000
         self.nMCMC = 10000 # N of sampling points
@@ -25,7 +26,7 @@ def __init__(self,data=[],response=[],background={},
         self.lower = lower  # lower sampling bounds
         self.upper = upper  # upper sampling bounds
         #                                     [unfolding model parameters]
-        self.prior = 'DiscreteUniform'
+        self.prior = 'Uniform'
         self.priorparams = {}
         self.regularization = regularization
         #                                     [input]
@@ -41,6 +42,9 @@ def __init__(self,data=[],response=[],background={},
         self.name      = name
         self.monitoring = monitoring
         self.sampling_progressbar = True
+        #                                     [mode]
+        self.mode = mode
+        self.MAP_method = 'L-BFGS-B'
 
     #__________________________________________________________
     def validateinput(self):
@@ -67,6 +71,7 @@ def run(self):
         nbckg = len(backgroundkeys)
 
         backgrounds = []
+        backgroundnormsysts = array([])
         if nbckg>0:
             backgrounds = array([self.background[key] for key in backgroundkeys])
             backgroundnormsysts = array([self.backgroundsyst[key] for key in backgroundkeys])
@@ -83,7 +88,7 @@ def run(self):
                                             for bckg in backgroundkeys])
 
         recodim  = len(data)
-        resmat = self.response
+        resmat   = self.response
         truthdim = len(resmat)
 
         model = mc.Model()
@@ -150,6 +155,18 @@ def unfold():
             import time
             from datetime import timedelta
             init_time = time.time()
+
+            print(self.nuts_kwargs)
+
+
+            if self.mode:
+                map_estimate = mc.find_MAP(model=model, method=self.MAP_method)
+                print (map_estimate)
+                self.MAP = map_estimate
+                self.trace = []
+                self.nuisancestrace = []
+                return
+
             trace = mc.sample(self.nMCMC,tune=self.nTune,cores=self.nCores,
                               chains=self.nChains, nuts_kwargs=self.nuts_kwargs,
                               discard_tuned_samples=self.discard_tuned_samples,
@@ -161,16 +178,20 @@ def unfold():
             ))
 
             self.trace = [trace['truth%d'%bin][:] for bin in range(truthdim)]
+            #self.trace = [copy.deepcopy(trace['truth%d'%bin][:]) for bin in range(truthdim)]
             self.nuisancestrace = {}
             if nbckg>0:
                 for name,err in zip(backgroundkeys,backgroundnormsysts):
                     if err<0.:
                         self.nuisancestrace[name] = trace['norm_%s'%name][:]
+                        #self.nuisancestrace[name] = copy.deepcopy(trace['norm_%s'%name][:])
                     if err>0.:
                         self.nuisancestrace[name] = trace['gaus_%s'%name][:]
+                        #self.nuisancestrace[name] = copy.deepcopy(trace['gaus_%s'%name][:])
             for name in objsystkeys:
                 if self.systfixsigma==0.:
                     self.nuisancestrace[name] = trace['gaus_%s'%name][:]
+                    #self.nuisancestrace[name] = copy.deepcopy(trace['gaus_%s'%name][:])
 
         if self.monitoring:
             from fbu import monitoring

setup.py

@@ -3,10 +3,10 @@
 DISTNAME = 'fbu'
 DESCRIPTION = "PyFBU"
 #VERSION = '0.0.2'
-VERSION = '0.1.1'
+VERSION = '0.1.2'
 AUTHOR = 'Davide Gerbaudo, Clement Helsens and Francesco Rubbo'
-AUTHOR_EMAIL = 'rubbo.francesco@gmail.com'
-URL = 'https://github.com/gerbaudo/fbu'
+AUTHOR_EMAIL = 'clement.helsens@gmail.com'
+URL = 'https://github.com/pyFBU/fbu'
 
 classifiers = [
     'Development Status :: 3 - Alpha',