Thinking about the similarities and differences between Latent Dirichlet Allocation $($LDA$),$ Non$-$Negative Matrix Factorization $($NMF$),$ Latent Semantic Indexing $($LSI$),$ and Principal Components Analysis $($PCA$),$ which of the following statements are TRUE? Select all that apply.

$1$ point

LDA and NMF can be seen as forms of matrix factorization, but LSI and PCA are not.

LSI and PCA are both based on Singular Value Decomposition.

LDA is a probabilistic method, while NMF is not a probabilistic method.

All four methods require the number of topics$/$latent dimensions$/$principal components to be specified in advance.

Methods like LDA and NMF are optimization$-$based methods with complex solution spaces, so different random initializations can lead to different results.