Merge branch 'master' of https://git.coding.net/wegamekinglc/Alpha-Mind

README.md

@@ -44,7 +44,9 @@ alpha - mind 的安装极其简单，只需要在下载源码之后，运行：
 python setup.py install
 ```
 
-* *注意事项*: 在Linux系统上,请确保gcc版本大于5
+* *注意事项*: 
+1. 在Linux系统上,请确保gcc版本大于4.8;
+2. 在libs下面提供了依赖的一些库的二进制文件。linux版本的是在一台具有两个intel cpu的docker虚机上面编译完成的。如果需要实现最佳的性能，建议用户在目标机器上编译相关依赖的库。依赖的库源码地址：[portfolio-optimizer](https://github.com/alpha-miner/portfolio-optimizer)
 
 ## 数据源
 

alphamind/cython/optimizers.pyx

@@ -8,14 +8,12 @@ Created on 2017-7-20
 
 cimport numpy as cnp
 import numpy as np
-from libcpp.vector cimport vector
-from libcpp cimport nullptr
 
 
 cdef extern from "lpoptimizer.hpp" namespace "pfopt":
     cdef cppclass LpOptimizer:
         LpOptimizer(int, int, double*, double*, double*, double*) except +
-        vector[double] xValue()
+        double* xValue()
         double feval()
         int status()
 
@@ -23,18 +21,20 @@ cdef extern from "lpoptimizer.hpp" namespace "pfopt":
 cdef class LPOptimizer:
 
     cdef LpOptimizer* cobj
+    cdef int n
+    cdef int m
 
     def __cinit__(self,
                  cnp.ndarray[double, ndim=2] cons_matrix,
                  double[:] lbound,
                  double[:] ubound,
                  double[:] objective):
-        cdef int n = lbound.shape[0]
-        cdef int m = cons_matrix.shape[0]
+        self.n = lbound.shape[0]
+        self.m = cons_matrix.shape[0]
         cdef double[:] cons = cons_matrix.flatten(order='C');
 
-        self.cobj = new LpOptimizer(n,
-                                    m,
+        self.cobj = new LpOptimizer(self.n,
+                                    self.m,
                                     &cons[0],
                                     &lbound[0],
                                     &ubound[0],
@@ -50,7 +50,13 @@ cdef class LPOptimizer:
         return self.cobj.feval()
 
     def x_value(self):
-        return np.array(self.cobj.xValue())
+        cdef cnp.ndarray[double, ndim=1] res = np.zeros(self.n)
+        cdef double* c_arr = self.cobj.xValue()
+
+        for i in range(self.n):
+            res[i] = c_arr[i]
+
+        return res
 
 
 cdef extern from "mvoptimizer.hpp" namespace "pfopt":
@@ -65,7 +71,7 @@ cdef extern from "mvoptimizer.hpp" namespace "pfopt":
                     double*,
                     double*,
                     double) except +
-        vector[double] xValue()
+        double* xValue()
         double feval()
         int status()
 
@@ -73,6 +79,8 @@ cdef extern from "mvoptimizer.hpp" namespace "pfopt":
 cdef class QPOptimizer:
 
     cdef MVOptimizer* cobj
+    cdef int n
+    cdef int m
 
     def __cinit__(self,
                  double[:] expected_return,
@@ -84,27 +92,27 @@ cdef class QPOptimizer:
                  double[:] cubound=None,
                  double risk_aversion=1.0):
 
-        cdef int n = lbound.shape[0]
-        cdef int m
+        self.n = lbound.shape[0]
+        self.m = 0
         cdef double[:] cov = cov_matrix.flatten(order='C')
         cdef double[:] cons
 
         if cons_matrix is not None:
-            m = cons_matrix.shape[0]
+            self.m = cons_matrix.shape[0]
             cons = cons_matrix.flatten(order='C');
 
-            self.cobj = new MVOptimizer(n,
+            self.cobj = new MVOptimizer(self.n,
                                         &expected_return[0],
                                         &cov[0],
                                         &lbound[0],
                                         &ubound[0],
-                                        m,
+                                        self.m,
                                         &cons[0],
                                         &clbound[0],
                                         &cubound[0],
                                         risk_aversion)
         else:
-            self.cobj = new MVOptimizer(n,
+            self.cobj = new MVOptimizer(self.n,
                                         &expected_return[0],
                                         &cov[0],
                                         &lbound[0],
@@ -122,7 +130,13 @@ cdef class QPOptimizer:
         return self.cobj.feval()
 
     def x_value(self):
-        return np.array(self.cobj.xValue())
+        cdef cnp.ndarray[double, ndim=1] res = np.zeros(self.n)
+        cdef double* c_arr = self.cobj.xValue()
+
+        for i in range(self.n):
+            res[i] = c_arr[i]
+
+        return res
 
     def status(self):
         return self.cobj.status()
\ No newline at end of file

libs/include/windows/pfopt/lpoptimizer.hpp

@@ -2,7 +2,6 @@
 #define pfopt_linear_programming_optimizer_hpp
 
 #include "types.hpp"
-#include <vector>
 #include "ClpSimplex.hpp"
 
 namespace pfopt {
@@ -16,14 +15,16 @@ namespace pfopt {
                     double* upperBound,
                     double* objective);
 
-        std::vector<double> xValue() const { return sol_; }
+        ~LpOptimizer() { delete [] sol_;}
+
+        double* xValue() const { return sol_; }
         double feval() const;
         int status() const { return model_.status(); }
 
     private:
         ClpSimplex model_;
         size_t numberOfProb_;
-        std::vector<double> sol_;
+        double* sol_;
     };
 }
 

libs/include/windows/pfopt/meanvariance.hpp

@@ -22,6 +22,8 @@ namespace pfopt {
                      double* varMatrix,
                      double riskAversion=1.);
 
+        virtual ~MeanVariance() { delete [] x_;}
+
         bool setBoundedConstraint(const double* lb, const double* ub);
         bool setLinearConstrains(int numCons, const double* consMatrix, const double* clb, const double* cub);
 
@@ -54,7 +56,7 @@ namespace pfopt {
                                        IpoptCalculatedQuantities *ip_cq);
 
         double feval() const { return feval_; }
-        std::vector<double> xValue() const { return x_; }
+        double* xValue() const { return x_; }
 
     private:
         VectorXd expectReturn_;
@@ -68,7 +70,7 @@ namespace pfopt {
         const double* ub_;
         VectorXd grad_f_;
         double feval_;
-        std::vector<double> x_;
+        double* x_;
         std::vector<Index> iRow_;
         std::vector<Index> jCol_;
         std::vector<double> g_grad_values_;

libs/include/windows/pfopt/mvoptimizer.hpp

@@ -18,7 +18,7 @@ namespace pfopt {
                     double *cub = nullptr,
                     double riskAversion = 1.);
 
-        std::vector<double> xValue() const { return mvImpl_->xValue(); }
+        double* xValue() const { return mvImpl_->xValue(); }
 
         double feval() const { return mvImpl_->feval(); }