Changeset 118

Timestamp:

12/22/07 22:16:40 (1 year ago)

Author:

edsuom

Message:

Usage example done

Files:

• projects/AsynCluster/trunk/doc/example.py (modified) (13 diffs)

projects/AsynCluster/trunk/doc/example.py

@@ -23 +23 @@
-"""
-This script shows how Bayesian inference can be run on a cluster of nodes to
-
+
+
-"""
-
@@ -32 +33 @@
-from scipy import stats
-
-
+, threads
-
-from asynqueue import ThreadQueue
@@ -166 +167 @@
-        return self.client.startup().addCallback(maybeStarted)
-    
-
+, runLocally=False
-        """
-        Returns the log-likelihood of my data given the Cauchy distribution of
@@ -176 +177 @@
-        example a 'non-informative uniform prior' is used.
-        """
+        def localLikelihood():
+            distObj = stats.distributions.cauchy(loc=X[0], scale=X[1])
+            return s.log(distObj.pdf(self.data)).sum()
+        
-        def gotResult(L_packed):
-            return pack.Unpacker(L_packed)()
-
-
-
-
+
+
+
+
+
+
+
-        return d
-
@@ -199 +207 @@
-    variance of the data set.
-    """
+    runLocally = False
-    socket="/tmp/.ndm"
-75, 0.3, 0.1, 0.025
+10, 0.02, 0.005, 0.001
-
-    def __init__(self, data):
@@ -246 +255 @@
-        # the most time-consuming step in real-life usage.
-        XD = self.proposer.r(len(X), v)
-
+, self.runLocally
-        d = defer.gatherResults(dList)
-        d.addCallback(gotLikelihoods)
@@ -252 +261 @@
-    
-    @defer.deferredGenerator
-bi, N_
+
-        """
-        Does a PMC run with I{N_chains} population members and jump variances
@@ -266 +275 @@
-        provided via the deferred, however.
-
-        Use a provider of L{interfaces.IConsumer} to get the samples as they
-        are produced. Each attached consumer is updated after every iteration
-        with my array of parameter vectors (one vector per row) for each
-        iteration.
-
-        @param N_iter: The number of iterations to produce after burn-in.
-
-        @param N_bi: The number of burn-in iterations to discard before
-          producing any.
-
-        @param N_chains: The number of Monte Carlo chains run at each
@@ -311 +312 @@
-        X = self.proposer.r(N_chains, self.V[0])
-        # The iteration loop
-
-
+
-            t0 = time.time()
-            # Generate and weight some proposals
@@ -329 +329 @@
-                X = s.row_stack(XP)[I]
-            R = s.array([sum([x in II[k] for x in I]) for k in xrange(P)])
-if i >= N_bi:
-    # Write the parameters for this iteration to the output file
-for Xk in X:
-    self.fh.write(", ".join(["%.3g" % x for x in Xk])
+# Write the parameters for this iteration to the output file
+for Xk in X:
+rowString = ", ".join(["%f" % x for x in Xk])
+self.fh.write(rowString + "\n"
-            # Normalize R to maintain a total of N_chains population members
-            normalize(R)
@@ -341 +341 @@
-                ", ".join(["%9.3g" % x for x in X.mean(0)]))
-            print msg
-            # Ready for the next iteration
-            i += 1
-        # All done
-        self.fh.close()
@@ -350 +348 @@
-if __name__ == '__main__':
-    # Parameters for the example
-
+0
-    N_chains = 500
-bi, N_iter = 100,
+iter =
-    loc, scale = 0.1, 0.4
-
@@ -358 +356 @@
-    distObj = stats.distributions.cauchy(loc=loc, scale=scale)
-    data = distObj.rvs(N_obs)
-
+
+
+
+
+
-    # Create the Monte Carlo runner and set up the run
-    pmc = Population_Monte_Carlo(data)
-bi, N_
+
-
-    # Run!