For a dataset with a million data points, would you use a DNN or KNN? Why?

For a dataset with a million data points, a Deep Neural Network (DNN) is generally preferred over K-Nearest Neighbors (KNN). KNN's primary drawback is its computational cost during prediction: it needs to compare the new data point to every point in the training set, which becomes prohibitively slow with a million data points. DNNs, on the other hand, learn a compressed representation of the data during training and can make predictions much faster, making them more scalable and efficient for large datasets, especially in high-dimensional spaces.