コード例 #1
0
ファイル: dualksvm2.py プロジェクト: percyqdeng/dualsvm
 def _fit_on_fly(self, xte, yte):
     if self.algo_type == 'cd':
         res = dsvm.coord_descent(xtr=self.xtr, ytr=self.ytr, kernel=self.kernel, xte=xte, yte=yte,
                                          verbose=self.verbose, lmda=self.lmda, nsweep=np.int(self.nsweep))
     elif self.algo_type == 'scg_da':
         res = dsvm.stocda_on_fly(xtr=self.xtr, ytr=self.ytr, kernel=self.kernel, xte=xte, yte=yte,
                                             lmda=self.lmda, rho=self.rho, verbose=np.int(self.verbose),
                                             nsweep=np.int(self.nsweep), b=np.int(self.b), c=np.int(self.c))
     self.alpha, self.err_tr, self.err_te, self.obj, self.nker_opers = res
コード例 #2
0
ファイル: dualksvm.py プロジェクト: percyqdeng/dualsvm
 def fit(self, xtr, ytr, xte=None, yte=None):
     self.xtr = xtr
     self.ytr = ytr
     self.construct_dataset(xtr, ytr, xte, yte)
     if self.algo_type == 'cd':
         res = dsvm.coord_descent(self.dataset, nsweep=np.int(self.nsweep), lmda=self.lmda, verbose=self.verbose)
         self.alpha, self.err_tr, self.err_te, self.obj, self.nker_opers = res
     elif self.algo_type == 'scg_da':
         res = dsvm.coord_dual_averaging(self.dataset, verbose=self.verbose, lmda=self.lmda, b=int(self.b),
                                         c=int(self.c), nsweep=np.int(self.nsweep), rho=self.rho)
         self.alpha, self.err_tr, self.err_te, self.obj, self.nker_opers = res
     elif self.algo_type == 'sbmd':
         res = dsvm.coord_mirror_descent(self.dataset, verbose=self.verbose, lmda=self.lmda, b=int(self.b),
                                         c=int(self.c), nsweep=np.int(self.nsweep), rho=self.rho)
         self.alpha, self.err_tr, self.err_te, self.obj, self.nker_opers, self.err_tr2, self.obj2= res
     else:
         raise NotImplementedError
コード例 #3
0
ファイル: dualksvm2.py プロジェクト: percyqdeng/dualsvm
    def _fit_precom(self, xte, yte):
        # ---------------precompute the kernel-----------------
        self.ktr = self.kernel_matrix(self.xtr)
        if xte is not None:
            self.kte = self.kernel_matrix(xte, self.xtr)
        if self.nsweep is None:
            if self.algo_type == 'scg_da':
                self.nsweep = self.xtr.shape[0]
            elif self.algo_type == 'cd':
                self.nsweep = self.xtr.shape[0]

        if self.algo_type == 'cd':
            res = dsvm.coord_descent(ktr=self.ktr, ytr=self.ytr, kte=self.kte, yte=yte, verbose=self.verbose,
                                 lmda=self.lmda, nsweep=np.int(self.nsweep))
        elif self.algo_type == 'scg_da':
            res = dsvm.stoc_dual_averaging(ktr=self.ktr, ytr=self.ytr, kte=self.kte, yte=yte, lmda=self.lmda, rho=self.rho,
                                      verbose=self.verbose, nsweep=np.int(self.nsweep), b=np.int(self.b), c=np.int(self.c))
        self.alpha, self.err_tr, self.err_te, self.obj, self.nker_opers = res