Updated Dependencies

.github/workflows/unit-tests.yml

@@ -16,7 +16,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.9", "3.10", "3.11"]
+        python-version: ["3.10", "3.11"]
 
     steps:
     - uses: actions/checkout@v3

xgboostlss/utils.py

@@ -1,6 +1,7 @@
 import torch
 from torch.nn.functional import softplus, gumbel_softmax, softmax
 
+
 def nan_to_num(predt: torch.tensor) -> torch.tensor:
     """
     Replace nan, inf and -inf with the mean of predt.
@@ -119,6 +120,71 @@ def softplus_fn_df(predt: torch.tensor) -> torch.tensor:
     return predt + torch.tensor(2.0, dtype=predt.dtype)
 
 
+def softplus_fn_quantile(predt: torch.tensor,
+                         beta: int,
+                         threshold: int) -> torch.tensor:
+    """
+    Softplus function used for Student-T distribution.
+
+    Arguments
+    ---------
+    predt: torch.tensor
+        Predicted values.
+    beta: int
+        Beta parameter for softplus function.
+    threshold: int
+        Threshold parameter for softplus function.
+
+    Returns
+    -------
+    predt: torch.tensor
+        Predicted values.
+    """
+    predt = softplus(input=nan_to_num(predt), beta=beta, threshold=threshold) + torch.tensor(1e-06, dtype=predt.dtype)
+
+    return predt
+
+
+def squareplus_fn(predt: torch.tensor) -> torch.tensor:
+    """
+    Square-Plus function used to ensure predt is strictly positive.
+
+    Arguments
+    ---------
+    predt: torch.tensor
+        Predicted values.
+
+    Returns
+    -------
+    predt: torch.tensor
+        Predicted values.
+    """
+    b = torch.tensor(4., dtype=predt.dtype)
+    predt = 0.5 * (predt + torch.sqrt(predt ** 2 + b)) + torch.tensor(1e-06, dtype=predt.dtype)
+
+    return predt
+
+
+def squareplus_fn_df(predt: torch.tensor) -> torch.tensor:
+    """
+    Square-Plus function used to ensure predt is strictly positive.
+
+    Arguments
+    ---------
+    predt: torch.tensor
+        Predicted values.
+
+    Returns
+    -------
+    predt: torch.tensor
+        Predicted values.
+    """
+    b = torch.tensor(4., dtype=predt.dtype)
+    predt = 0.5 * (predt + torch.sqrt(predt ** 2 + b)) + torch.tensor(1e-06, dtype=predt.dtype)
+
+    return predt + torch.tensor(2.0, dtype=predt.dtype)
+
+
 def sigmoid_fn(predt: torch.tensor) -> torch.tensor:
     """
     Function used to ensure predt are scaled to (0,1).
@@ -158,6 +224,25 @@ def relu_fn(predt: torch.tensor) -> torch.tensor:
     return predt
 
 
+def relu_fn_df(predt: torch.tensor) -> torch.tensor:
+    """
+    Function used to ensure predt are scaled to max(0, predt).
+
+    Arguments
+    ---------
+    predt: torch.tensor
+        Predicted values.
+
+    Returns
+    -------
+    predt: torch.tensor
+        Predicted values.
+    """
+    predt = torch.relu(nan_to_num(predt)) + torch.tensor(1e-06, dtype=predt.dtype)
+
+    return predt + torch.tensor(2.0, dtype=predt.dtype)
+
+
 def softmax_fn(predt: torch.tensor) -> torch.tensor:
     """
     Softmax function used to ensure predt is adding to one.
@@ -216,5 +301,4 @@ def gumbel_softmax_fn(predt: torch.tensor,
     torch.manual_seed(123)
     predt = gumbel_softmax(nan_to_num(predt), tau=tau, dim=1) + torch.tensor(0, dtype=predt.dtype)
 
-
     return predt