Neural Tangent Kernel
From the last post, we know that a wide neural network with reasonable initialization could be seen as a Gaussian Porcess without training. After training, what can we say about wide NN?
阅读全文From the last post, we know that a wide neural network with reasonable initialization could be seen as a Gaussian Porcess without training. After training, what can we say about wide NN?
阅读全文Interesting Idea…
阅读全文很优美的计算方法。
继续阅读→
雅礼集训某天的东西。。
继续阅读→