A company plans to use a SageMaker built-in algorithm to build a recommender that must predict on high-dimensional, sparse input data. Which SageMaker algorithm is appropriate for this recommendation problem?

Factorization Machines (FM) are well-suited for recommender systems, especially with high-dimensional, sparse data, because they can capture interactions between features even when those interactions are rarely observed together. This makes them effective for predicting user preferences based on limited historical data. K-nearest neighbors (k-NN) is a non-parametric algorithm that can struggle with very high-dimensional data due to the "curse of dimensionality" and is less efficient with sparse inputs for recommendation tasks. Principal component analysis (PCA) is primarily a dimensionality reduction technique, not a recommendation algorithm, and wouldn't directly predict recommendations. Sequence-to-Sequence (seq2seq) models are designed for tasks involving sequences, like machine translation or time-series prediction, not typically for static, sparse recommender systems.