Which tool is used to explore alternative network architectures and hidden unit counts?

The AutoNeural tool is specifically designed for exploring alternative neural network architectures and optimizing the number of hidden units in a model. This tool automates the process of adjusting various configurations, allowing practitioners to experiment with different network setups without the need for manual tuning of parameters. Using AutoNeural streamlines the workflow and helps identify the most effective network configuration for a given dataset.

In contrast, the Neural Network tool allows users to build and train a neural network but does not inherently provide the same level of automation for testing multiple architectures or hidden unit counts. Data Partitioning is focused on dividing data into training, validation, and testing sets rather than optimizing neural network structures. Lastly, Model Comparison is aimed at evaluating the performance of different models against each other rather than exploring architectural variations. Thus, AutoNeural stands out as the best choice for tasks involving the exploration of alternative architectures and hidden units in neural network modeling.