Differences

This shows you the differences between two versions of the page.

--- personal:portfolio:portopt [2014/06/06 10:22]
antonello [Usage]
+++ personal:portfolio:portopt [2015/09/10 15:19]
antonello [Compilation (not needed if using a pre-compiled version)]
@@ Line 54: / Line 54: @@
 \end{equation}
-where $x_i$ is the share of the asset $i$, $p_i$ is its productivity and hence $\sum_i {x_i p_i}$ is the overall portfolio productivity and $\sum_i { \sum_j { x_i x_j \sigma_{i,j}}}$ its variance.
+where $x_i$ is the share of the asset $i$, $p_i$ is its productivity, $\sigma_{i,j}$ is the covariance between assets $i$ and $j$ and hence $\sum_i {x_i p_i}$ is the overall portfolio productivity and $\sum_i { \sum_j { x_i x_j \sigma_{i,j}}}$ is its variance.
 As the only quadratic term arises when $i=j$ and $\sigma_{i,j}$ being the variance is always positive, the problem is convex and hence easily numerically solved.
@@ Line 81: / Line 81: @@
   g++ -std=c++0x -O -shared -Wl,-soname,_portopt.so -o _portopt.so QuadProg++.o Array.o anyoption.o portopt.o portopt_wrap.o
 (then please refer to the python example for usage)
+If you want to change the output library name (e.g. you want to create _portopt_p3.so for python3 alongside _portopt.so for python2), do it in the %module variable of portopt.i and in the -soname and -o options of the linking command (and don't forget to use the right python included directory in the compilation command).\\
+You can then load the correct module in your script with something like:
+  import sys
+  if sys.version_info < (3, 0):
+    import portopt
+  else:
+    import portopt_p3 as portopt
 ===== Usage =====
@@ Line 97: / Line 105: @@
 == As a lib using Python: ==
   import portopt
-  results = portopt.solveport(var,means,alpha,tollerance) # tollerance is optional, default to 0.000001
+  results = portopt.solveport(var,means,alpha,tolerance) # tolerance is optional, default to 0.000001
   functioncost = results[0]
   shares       = results[1]
@@ Line 120: / Line 128: @@
   * Set alpha to zero to retrieve the portfolio with the highest mean, independently from variance (solution not guaranteed to be unique);
   * Assets shares are returned in the x_h vector, eventual error code (0: all fine, 1: input data error, 2: no solutions, 3: didn't solve, 4: solver internal error) in the errorcode parameter.
-  * Use option "tollerance" with two l up to version 1.1
+  * Use option "tollerance" with two l up to version 1.1 included