General matrix multiply for ndarrays. This is analogous to the BLAS level 3 routine xGEMM.
Note that while this implementation is correct, it is not yet very optimized. If someone wants to take over this project or suggest improvements, patches are welcome.
var zeros = require("zeros")
var ops = require("ndarray-ops")
var gemm = require("ndarray-gemm")
//Create 3 random matrices
var a = zeros([300, 400]) // a is 300 x 400
var b = zeros([400, 500]) // b is 400 x 500
var c = zeros([300, 500]) // c is 300 x 500
ops.random(a)
ops.random(b)
ops.random(c)
//Set c = a * b
gemm(c, a, b)
Install using npm:
npm install ndarray-gemm
require("ndarray-gemm")(c, a, b[, alpha, beta])
Computes a generalized matrix multiplication. This sets:
c = alpha * a * b + beta * c
c
is a [n,m]
shape ndarraya
is a [n,p]
shape ndarrayb
is a [p,m]
shape ndarrayalpha
is a scalar weight which is applied to the product a * b
beta
is a scalar weight applied to c
when added back in(c) 2013 Mikola Lysenko. MIT License