Abstract: Neural net compression can be achieved by approximating each layer's weight matrix by a low-rank matrix. The real difficulty in doing this is not in training the resulting neural net (made ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback