Spare random projection (SRP) is a great technique in terms of speed and also in terms of computational sources.
Because SPR is a way of doing matrix multiplication, it doesn't require Computation and doesn't "consume a lot of RAM" so it's a good choice in general without worrying about the computational resource.
But if we're talking about accuracy, I am not sure that SPR is a good choice here, and I would really prefer using PCA instead of it. Since SPR is based on a randomized algorithm, its output can vary with the same input. While PCA is a deterministic algorithm, it will always give the same output for the same input.
Finally, the dimension reduction technique has a lot of in-depth details to go in.
Well done mate! 👏
Spare random projection (SRP) is a great technique in terms of speed and also in terms of computational sources.
Because SPR is a way of doing matrix multiplication, it doesn't require Computation and doesn't "consume a lot of RAM" so it's a good choice in general without worrying about the computational resource.
But if we're talking about accuracy, I am not sure that SPR is a good choice here, and I would really prefer using PCA instead of it. Since SPR is based on a randomized algorithm, its output can vary with the same input. While PCA is a deterministic algorithm, it will always give the same output for the same input.
Finally, the dimension reduction technique has a lot of in-depth details to go in.