This is burying the lede a bit, which is that with a minor change you can also do ReLU(A @ B) @ C on NxN matrices with N intermediates and no recomputation. (aka: a neural network.)
Matrix Operations Optimization for Neural Networks
By
–
By
–
This is burying the lede a bit, which is that with a minor change you can also do ReLU(A @ B) @ C on NxN matrices with N intermediates and no recomputation. (aka: a neural network.)