Because clustering doesn't produce or include a ground "truth" against which you can verify the output, it's important to check the result against your expectations at both the cluster level and the example level. If the result looks odd or low-quality, experiment with the previous three steps. Continue iterating until the quality of the output meets your needs.

A clustering algorithm uses the similarity metric to cluster data. This course uses k-means.

Before a clustering algorithm can group data, it needs to know how similar pairs of examples are. You can quantify the similarity between examples by creating a similarity metric, which requires a careful understanding of your data.

As with any ML problem, you must normalize, scale, and transform feature data before training or fine-tuning a model on that data. In addition, before clustering, check that the prepared data lets you accurately calculate similarity between examples.

Centroid-based clustering

The k-means algorithm has a complexity of O(n)O(n), meaning that the algorithm scales linearly with nn. This algorithm will be the focus of this course.

clustering algorithms compute the similarity between all pairs of examples, which means their runtime increases as the square of the number of examples nn, denoted as O(n2)O(n^2) in complexity notation

Clustering YouTube videos replaces this set of features with a single cluster ID, thus compressing the data.

As discussed, the relevant cluster ID can replace other features for all examples in that cluster. This substitution reduces the number of features and therefore also reduces the resources needed to store, process, and train models on that data. For very large datasets, these savings become significant.

When some examples in a cluster have missing feature data, you can infer the missing data from other examples in the cluster.

After clustering, each group is assigned a unique label called a cluster ID. Clustering is powerful because it can simplify large, complex datasets with many features to a single cluster ID.

Different similarity measures may be more or less appropriate for different clustering scenarios, and this course will address choosing an appropriate similarity measure in later sections: Manual similarity measures and Similarity measure from embeddings.

But as the number of features increases, combining and comparing features becomes less intuitive and more complex.

similarity measure, or the metric used to compare samples,

Clustering is an unsupervised machine learning technique designed to group unlabeled examples based on their similarity to each other. (If the examples are labeled, this kind of grouping is called classification.)

Tables are an intuitive input format for machine learning models. You can imagine each row of the table as an example and each column as a potential feature or label. That said, datasets may also be derived from other formats, including log files and protocol buffers.

Many datasets store data in tables (grids), for example, as comma-separated values (CSV) or directly from spreadsheets or database tables.

Yes, ML practitioners spend the majority of their time constructing datasets and doing feature engineering.

Data trumps all. The quality and size of the dataset matters much more than which shiny algorithm you use to build your model.