[$] Deep learning and free software

<p>Deep-learning applications typically rely on a trained neural net toaccomplish their goal (e.g. photo recognition, automatic translation, orplaying go). That neural net uses what is essentially a large collection ofweighting numbers that have been empirically determined as part of its training (which generally uses a huge set of training data). Afree-software application could use those weights, but there are a number of barriers for users who might want to tweak them for variousreasons. A discussion on the debian-devel mailing list recently looked atwhether these deep-learning applications can ever truly be considered "free" (as infreedom) because of these pre-computed weights-and the difficultiesinherent in changingthem.