Reference:
https://cloud.google.com/ai-platform/training/docs/using-hyperparameter-tuning#early-stopping

- **Categorical entropy** is better to use when you want to **prevent the model from giving more importance to a certain class**. Or if the **classes are very unbalanced** you will get a better result by using Categorical entropy.
- But **Sparse Categorical Entropy** is a more optimal coice if you have a huge amount of classes, enough to make a lot of memory usage, so since sparse categorical entropy uses less columns it **uses less memory**.
https://stats.stackexchange.com/questions/326065/cross-entropy-vs-sparse-cross-entropy-when-to-use-one-over-the-other

https://cloud.google.com/ai-platform/training/docs/machine-types#scale_tiers Google may optimize the configuration of the scale tiers for different jobs over time, based on customer feedback and the availability of cloud resources. Each scale tier is defined in terms of its suitability for certain types of jobs. Generally, the more advanced the tier, the more machines are allocated to the cluster, and the more powerful the specifications of each virtual machine. As you increase the complexity of the scale tier, the hourly cost of training jobs, measured in training units, also increases. See the pricing page to calculate the cost of your job.