added bias to convolution, improved init of parameters

src/UQpy/scientific_machine_learning/layers/Fourier1d.py

@@ -62,7 +62,7 @@ def __init__(
         :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})` 
         where :math:`k = \frac{1}{\text{width}}`"""
         self.bias_conv: nn.Parameter = nn.Parameter(
-            torch.empty(self.width, self.width, kernel_size, device=device)
+            torch.empty(self.width, device=device)
         )
         r"""The learnable bias of the convolution of shape :math:`(\text{width})`.
         
@@ -71,7 +71,7 @@ def __init__(
         where :math:`k = \frac{1}{\text{width}}`
         """
 
-        k = torch.sqrt(1 / width)
+        k = torch.sqrt(1 / torch.tensor(self.width, device=device))
         self.reset_parameters(-k, k)
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:

src/UQpy/scientific_machine_learning/layers/Fourier2d.py

@@ -64,7 +64,7 @@ def __init__(
         )
         r"""The learnable bias of the convolution of shape :math:`(\text{width})`."""
 
-        k = torch.sqrt(1 / self.width)
+        k = torch.sqrt(1 / torch.tensor(self.width, device=device))
         self.reset_parameters(-k, k)
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:

src/UQpy/scientific_machine_learning/layers/Fourier3d.py

@@ -49,7 +49,7 @@ def __init__(
         :math:`(\text{width}, \text{width}, \text{modes[0]}, \text{modes[1]}, \text{modes[2]})` 
         with complex entries"""
         self.weight_spectral_2: nn.Parameter = nn.Parameter(
-            self.scale * torch.rand(shape, dtype=torch.cfloat, device=device)
+            torch.empty(shape, dtype=torch.cfloat, device=device)
         )
         r"""The second of four learnable weights for the spectral convolution of shape
         :math:`(\text{width}, \text{width}, \text{modes[0]}, \text{modes[1]}, \text{modes[2]})` 
@@ -76,7 +76,7 @@ def __init__(
             torch.empty(self.width, device=device)
         )
         r"""The learnable bias of the convolution of shape :math:`(\text{out_channels})`"""
-        k = torch.sqrt(1 / self.width)
+        k = torch.sqrt(1 / torch.tensor(self.width, device=device))
         self.reset_parameters(-k, k)
 
     def forward(self, x: torch.Tensor) -> torch.Tensor: