A data scientist needs to build a SageMaker built-in model to classify various vegetables. The dataset has many features and the company wants to minimize memory usage during training and deployment. The company also requires the ability to find similar data points for any test example. Which algorithm choice satisfies these constraints?

k-Nearest Neighbors (k-NN) is a suitable choice because it can classify data points and inherently supports finding similar data points (its core mechanism). Combining it with dimensionality reduction (e.g., PCA) addresses the memory constraint by reducing the number of features, making the model more efficient for training and deployment, especially with many features. Linear Learner is a classification algorithm but does not inherently support finding similar data points. K-means clustering is an unsupervised algorithm used for grouping data, not classification, and doesn't directly provide similarity search for new data points in the same way k-NN does. PCA is a dimensionality reduction technique, not a classification algorithm itself.