Resetting a notebook's state requires terminating the current kernel session, which clears all in-memory variables and execution history. This can be accomplished in several ways within the Azure Machine Learning Studio notebook interface.
1. Stopping the current kernel (C) or restarting it is the most direct method. This terminates the running process, and a new, clean kernel starts upon the next cell execution.
2. Changing the current kernel (B), for example, from a Python 3.8 kernel to a Python 3.10 kernel, also terminates the old session and starts a new one.
3. Changing the compute instance (D) the notebook is attached to will disconnect from the kernel on the original compute and start a fresh kernel session on the new compute.

A. Resetting the compute restarts the entire virtual machine. While effective, it is an overly broad action that unnecessarily affects all other sessions on that compute instance.
E. Navigating to another section of the workspace does not affect the kernel. The session persists on the compute instance, and its state is preserved upon your return.

1. Microsoft Learn. (2023). Run Jupyter notebooks in your workspace. "To change the compute that the notebook is running on, select a different compute from the Compute dropdown in the notebook toolbar." This directly supports option D.

2. Microsoft Learn. (2023). Run Jupyter notebooks in your workspace. "To change the kernel, select a different kernel from the Kernel dropdown in the notebook toolbar." This directly supports option B.

3. Microsoft Learn. (2023). Run Jupyter notebooks in your workspace. The documentation describes the Kernel menu, which includes options like Shutdown and Restart. Stopping the kernel (Shutdown) is a direct way to reset the state. This supports option C.

4. Microsoft Learn. (2023). How to create and manage an Azure Machine Learning compute instance. This document describes restarting a compute instance as a management action for the entire resource, often for updates or troubleshooting, confirming it is not a targeted action for a single notebook's state. This supports why option A is less appropriate.