python-ldl 0.1.0__py3-none-any.whl → 0.1.1__py3-none-any.whl

pyldl/algorithms/_duo_ldl.py

@@ -25,9 +25,9 @@ class Duo_LDL(BaseAdam, BaseDeepLDL):
         , axis=1)
 
     def predict(self, X):
-        from pyldl.algorithms.utils import normalize
+        from pyldl.algorithms.utils import proj
         C_pred = self._call(X)
         shape = (X.shape[0], self._n_outputs - 1, self._n_outputs)
         C_pred_reshaped = ops.transpose(ops.reshape(C_pred, shape), (0, 2, 1))
         D_pred = self._to_numpy((ops.sum(C_pred_reshaped, axis=2) + 1) / self._n_outputs)
-        return normalize(D_pred)
+        return proj(D_pred)

pyldl/algorithms/_ldl_dpa.py

@@ -36,4 +36,4 @@ class LDL_DPA(BaseBFGS, BaseDeepLDL):
 
     def _before_train(self):
         from scipy.stats import rankdata
-        self._R = ops.cast(rankdata(self._D, axis=1), dtype="float32")
+        self._R = ops.cast(rankdata(self._to_numpy(self._D), axis=1), dtype="float32")

pyldl/algorithms/_lrldl.py

@@ -11,8 +11,8 @@ class _LRLDL(BaseADMM, BaseLDL):
     :term:`ADMM` is used as optimization algorithm.
     """
 
-    def __init__(self, mode='threshold', param=None, alpha=1e-3, beta=1e-3, random_state=None, **kwargs):
-        super().__init__(random_state, **kwargs)
+    def __init__(self, mode='threshold', param=None, alpha=1e-3, beta=1e-3, **kwargs):
+        super().__init__(**kwargs)
         self._mode = mode
         self._param = param
         self._alpha = alpha