Machine Learning with Swift

Research the data

If you have not found an existing solution and want to train your own algorithm, you will need a dataset.

Here, several scenarios are possible:

• You are feeling lucky, and you have found an existing dataset. The potential problem is that you may not be the only lucky person and your approach can be copied by others. There can also be licensing issues or other related problems.
• You collect or generate your own dataset.
• In the case of supervised learning, your dataset should be labeled. Hand-labeling is a laborious task, so it is often outsourced to some third-party services, such as Amazon Mechanical Turk.

Calculate how much will it cost to label your data in terms of time and money.

Another important thing to mention is ...