📄 interpolate.html
字号:
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"><html><head></head><body><div style="text-align: center;"><big style="font-weight: bold;">MachineLerner Interpolation Tool</big><br>This uses several machine learning algorithms to blow up a small image.<br></div><br><table style="width: 100%; text-align: left;" border="0" cellpadding="2" cellspacing="2"> <tbody> <tr> <td style="vertical-align: top;">Note that there are much betteralgorithms for doing high-quality interpolation than any of thesemethods. The real purpose of this demo is to give a visual "feel" forthe differences between several learning algorithms.<br> <br>The tiny image in the top-left is the training set. Each algorithm hastwo inputs (x, y) and one output (grayscale value). The big images areproduced by evaluating with sub-pixel inputs.<br> <br>The four algorithms on the bottom are fast enough that theinterpolation is performed entirely before this screen appears.<br> <br>The six algorithms on the top all use the same neural net model, butuse different search algorithms to train it. You can watch in real timewhile they train. They each take turns so that each algorithm gets thesame amount of time in the long run. As you'll see, on-linebackpropagation beats the other algorithms. Momentum Greedy comes insecond. The others are painfully slow. I'm a bit uncertain if myPartical Swarm implementation is correct, because I'm told it is aneffective way to train a neural net, but it doesn't seem to work sowell for me.<br> </td> <td style="vertical-align: top; text-align: right;"><img alt="" src="Interpolate.png" style="width: 509px; height: 311px;"><br> </td> </tr> </tbody></table><br></body></html>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -