Changeset 733

Timestamp:

07/28/14 08:40:55 (22 months ago)

Author:

mmckerns

Message:

moved pyre examples to examples_other; cleaned up imports on scemtools

Location:

mystic

Files:

• examples/README (modified) (2 diffs)
• examples_other/chebyshevinputs.py (moved) (moved from mystic/tests/chebyshevinputs.py)
• examples_other/chebyshevinputs_de.py (moved) (moved from mystic/tests/chebyshevinputs_de.py)
• examples_other/derun.py (moved) (moved from mystic/examples/derun.py)
• examples_other/dummy.py (moved) (moved from mystic/examples/dummy.py)
• examples_other/roseninputs.py (moved) (moved from mystic/tests/roseninputs.py)
• examples_other/testsolvers_pyre.py (moved) (moved from mystic/tests/testsolvers_pyre.py)
• mystic/scemtools.py (modified) (9 diffs)

mystic/examples/README

@@ -1 +1 @@
 == Notes on mystic examples ==
-NOTE: for all examples that use matplotlib, please use the TKAgg backend.
-Thus, run the examples like this:  "python example04.py -dTKAgg"
-(see ticket #36 for more details).
+Dependencies:
+ - Several of the examples require matplotlib to be installed.
+ - For the examples that use matplotlib, see trac ticket #36 for more details.
 
-Dependencies:
- - All examples with prefix "example" should run without new dependencies, and are intended as a tutorial (i.e. TRY THESE FIRST).
+Other dependencies:
+ - Examples with prefix "example" are part of the tutorial (TRY THESE FIRST).
+ - All examples with prefix "example" should run without new dependencies.
  - All examples with prefix "test_" should run without new dependencies.
- - All examples with prefix "gplot_" requres gnuplot-py to be installed.
+ - All examples with prefix "gplot_" requre gnuplot-py for visualization.
 
 Exceptions to the rule:
- - The following examples also require scipy to be installed. (tested on version 0.4.8 and 0.7.0):
-    . test_lorentzian.py
-    . test_mogi_anneal.py,
-    . test_twistedgaussian.py
-    . test_twistedgaussian2.py
+ - The following examples also require scipy to be installed:
+   . test_lorentzian.py
+   . test_mogi_anneal.py,
 
 Special examples:
  - All examples with prefix "rosetta_" require park to be installed. (tests on version park-1.2).  Run with "--park" to execute with park. See "--help" for more options.
- - The derun.py example requires pyre (from pythia-0.8) to be installed, and drives the otherwise useless 'dummy.py'.
 
 
@@ -33 +31 @@
 Notes on the "mogi" tests/examples:
  - test_mogi.py: One mogi source with noise, comparison between DE and Conjugate Gradient, Simplex, and least squares (Levenberg Marquardt). CG / lsq don't work very well. lsq should work when bounds on the parameters are given, but minpack (wrapped by scipy) version doesn't seem to support bounds.
- - sam_mogi.py: charts the progress of scipy's Nelder Mead and draws with matlab
  - test_mogi_anneal.py: tests with scipy simulated annealing, but hasn't been tuned, so again, doesn't work at all.
  - test_mogi2.py: two mogi sources

mystic/mystic/scemtools.py

@@ -26 +26 @@
 """
 
-from numpy import mean, cov, array, zeros, identity, transpose
-from numpy.random import multivariate_normal
 import numpy
 try:
@@ -39 +37 @@
 def multinormal_pdf(mean, var):
     """var must be symmetric positive definite """
-    from numpy.linalg import inv, det
-    from numpy import transpose, array, exp, sqrt, pi, dot
-    vinv = inv(var)
-    mu = array(mean).flatten()
+    import numpy
+    vinv = numpy.linalg.inv(var)
+    mu = numpy.array(mean).flatten()
     n = len(mu)
 
     # check that var is properly sized
-    dum = dot(mu,var) - dot(var,mu)
-
-    prefactor = 1./sqrt((2 * pi)**n  * det(var))
+    dum = numpy.dot(mu,var) - numpy.dot(var,mu)
+
+    prefactor = 1./numpy.sqrt((2 * numpy.pi)**n  * numpy.linalg.det(var))
     def _(x):
-        xm = array(x) - mu
-        return prefactor *  exp(-0.5 * dot(xm,dot(vinv,xm)) )
+        xm = numpy.array(x) - mu
+        return prefactor *  numpy.exp(-0.5 * numpy.dot(xm,numpy.dot(vinv,xm)) )
     return _
 
@@ -76 +73 @@
 
     """
-    from numpy import array, transpose
-    cards = array(inarray)
+    import numpy
+    cards = numpy.array(inarray)
     N = len(cards)
     # this bit of numpy will give, for N=20, n = 5
     # ord = [ [0,5,10,15], [1,6,11,16], [2,7,12,17], [3,8,13,18], [4,9,14,19] ]
-    ord = transpose(array(range(N)).reshape(N/n, n))
+    ord = numpy.transpose(numpy.array(range(N)).reshape(N/n, n))
     return [cards[x] for x in ord] 
 
 def sort_and_deal(cards, target, nplayers):
-    from numpy import argsort
-    c = array(map(target, cards))
-    o = list(reversed(argsort(c)))
+    import numpy
+    c = numpy.array(map(target, cards))
+    o = list(reversed(numpy.argsort(c)))
     # from best to worst
     sorted_deck = cards[o]
@@ -95 +92 @@
     """default is descending..."""
     import numpy
-    aa,bb = array(a), array(b)
+    aa,bb = numpy.array(a), numpy.array(b)
     if ord == -1:
         o = list(reversed(bb.argsort()))
@@ -106 +103 @@
     # should use the one below instead.
     # this is faster than sort_complex
-    b = array(a)
+    import numpy
+    b = numpy.array(a)
     o = list(reversed(b.argsort()))
     return c[o], list(b[o])
@@ -113 +111 @@
     # this is dumb, because c (i.e., a, are almost sorted)
     # should use the one below instead.
+    import numpy
     D = zip(a,c)
     def mycmp(x,y):
@@ -201 +200 @@
 
     """
-
+    import numpy
     # function level constants
     T = 100000. # predefined likelihood ratio. Paragraph 45 of [1.]
 
-    from numpy import mean, cov, array, zeros, identity, transpose
-    
     # Sort Ck according to ak
     #Ck, ak = sort_complex0(Ck, ak) 
@@ -217 +214 @@
     M = Ck.shape[0]
 
-    mu = mean(Ck,0) # row mean
+    mu = numpy.mean(Ck,0) # row mean
 
     # (numpy.cov takes data in columns)
-    Sigma = cov(transpose(Ck)) 
+    Sigma = numpy.cov(numpy.transpose(Ck)) 
 
     # Gamma (line 35 of [1]. Best to Worst)
@@ -226 +223 @@
    
     if len(Sak) >= M:
-        meansak = mean(Sak[-M:])
-    else:
-        meansak = mean(Sak)
-    alpha_k = mean(ak) / (meansak+TINY)
+        meansak = numpy.mean(Sak[-M:])
+    else:
+        meansak = numpy.mean(Sak)
+    alpha_k = numpy.mean(ak) / (meansak+TINY)
   
     if alpha_k < T:
          # Paragraph 37 of [1]
          basept = Sk[-1]
-    else: # mean(asak) is very close to zero.
+    else: # numpy.mean(asak) is very close to zero.
          # Paragraph 38 of [1]
          basept = mu
 
     # TODO should take a proposal instead !
-    Yt = multivariate_normal(basept,  cn*cn * Sigma)
+    Yt = numpy.random.multivariate_normal(basept,  cn*cn * Sigma)
     cY = target(Yt)