Transfer Learning Is Really Good for Fast Iteration Time

Geridofish: We tried the dumbest thing and it worked. Transfer learning is really good for fast iteration time, which i think is really crucial for research. I have talked to some people at other la. They read an experiment on their laps, like sherd cluster, and fight for time. It's pretty painful to get the bug andm ex wis types of problems. The devile works. And if you were to train from scratch, i'm sure it would work too. Wouldnt be half done but it's just exen to do.