[Term Entry] PyTorch Tensor Operations: .signbit()

content/pytorch/concepts/tensor-operations/terms/signbit/signbit.md

@@ -0,0 +1,116 @@
+---
+Title: '.signbit()'
+Description: 'Returns a boolean tensor indicating whether each element has its sign bit set.'
+Subjects:
+  - 'Computer Science'cod
+  - 'Data Science'
+  - 'Machine Learning'
+Tags:
+  - 'Integers'
+  - 'Python'
+  - 'Tensor'
+CatalogContent:
+  - 'intro-to-py-torch-and-neural-networks'
+  - 'paths/computer-science'
+---
+
+The **`.signbit()`** function in PyTorch checks the sign bit of each element in a tensor and returns a boolean tensor. It evaluates to `true` for values that are negative (including -0.0) and `false` otherwise.
+
+Unlike simple comparison operators, `.signbit()` works at the numerical representation level, making it useful when dealing with edge cases such as negative zero and floating-point sign handling.
+
+## Syntax
+
+```pseudo
+torch.signbit(input) → Tensor
+```
+
+**Parameters:**
+
+- `input`: A tensor containing numeric values whose sign bits are to be checked.
+
+**Return value:**
+
+Returns a boolean tensor of the same shape as `input`, where:
+
+- `true` indicates the value has its sign bit set
+- `false` indicates the value does not have its sign bit set
+
+## Example 1: Detecting Negative Values
+
+This example demonstrates how `.signbit()` identifies negative numbers in a tensor:
+
+```py
+import torch
+
+x = torch.tensor([-3.0, -1.0, 0.0, 2.5])
+
+result = torch.signbit(x)
+
+print("Input tensor:", x)
+print("Sign bit result:", result)
+```
+
+The output of this code is:
+
+```shell
+Input tensor: tensor([-3.0000, -1.0000,  0.0000,  2.5000])
+Sign bit result: tensor([ True,  True, False, False])
+```
+
+Values less than zero return `true`, while zero and positive values return `false`.
+
+## Example 2: Handling Negative Zero (`-0.0`)
+
+This example shows how `.signbit()` treats negative zero differently from positive zero:
+
+```py
+import torch
+
+x = torch.tensor([0.0, -0.0, 1.0, -1.0])
+
+result = torch.signbit(x)
+
+print("Input tensor:", x)
+print("Sign bit result:", result)
+```
+
+The output of this code is:
+
+```shell
+Input tensor: tensor([ 0., -0.,  1., -1.])
+Sign bit result: tensor([False,  True, False,  True])
+```
+
+Here, `-0.0` returns `true` because its sign bit is set, even though it compares equal to `0.0`.
+
+## Example 3: Multi-Dimensional Tensor
+
+This example applies `.signbit()` to a 2D tensor:
+
+```py
+import torch
+
+x = torch.tensor([[1.0, -2.0],
+                  [-0.0, 3.5]])
+
+result = torch.signbit(x)
+
+print("Input tensor:")
+print(x)
+print("\nSign bit result:")
+print(result)
+```
+
+The output of this code is:
+
+```shell
+Input tensor:
+tensor([[ 1.0000, -2.0000],
+        [-0.0000,  3.5000]])
+
+Sign bit result:
+tensor([[False,  True],
+        [ True, False]])
+```
+
+The returned tensor preserves the original shape and indicates the sign bit status for each element.