Fix polynomial_autograd documentation to match implementation

The tutorial code uses exp(x) but the documentation incorrectly stated
it was learning sin(x). This fix updates the documentation to correctly
describe what the code does:

- Changed documentation from sin(x) from -π to π
- To exp(x) from -1 to 1
- Added explanation about Taylor expansion of exponential function

The code remains unchanged and continues to demonstrate fitting a
polynomial to the exponential function using PyTorch autograd.

Author: patrocinio

beginner_source/examples_autograd/polynomial_autograd.py

@@ -2,8 +2,9 @@
 PyTorch: Tensors and autograd
 -------------------------------
 
-A third order polynomial, trained to predict :math:`y=\sin(x)` from :math:`-\pi`
-to :math:`\pi` by minimizing squared Euclidean distance.
+A third order polynomial, trained to predict :math:`y=e^x` from :math:`-1`
+to :math:`1` by minimizing squared Euclidean distance. The exponential function
+can be approximated by its Taylor expansion: :math:`e^x \approx 1 + x + \frac{x^2}{2} + \frac{x^3}{6} + \ldots`
 
 This implementation computes the forward pass using operations on PyTorch
 Tensors, and uses PyTorch autograd to compute gradients.