From 2011-2012, the FTC investigated the possibility of pursuing antitru…

Ciprian Demeter, Michael Lacey, Christoph Thiele and I have just uploade…

用 6 张 matplotlib 图和一个真实可运行的 toy MLP，把神经网络从单神经元、前向传播、损失函数、反向求导、梯度下降、NumPy…

We’re going to reproduce some results from papers on filesystem robustne…

This post will have only the most tangential connection to mathematics. …

Softmax 不是一个孤立的归一化函数，而是把任意实数分数变成概率分布的一座桥。本文从’为什么需要它’出发，讲清楚公式、几何、温度、稳定性、与…

I haven’t used a desktop email client in years. None of them could handl…

Next quarter, starting on Wednesday January 9, I will be teaching a grad…

神经网络真正会「学习」靠的是两件事：把误差变成可微分的损失函数，再沿着这个损失对参数的梯度方向一点点往下挪。本文从一维抛物线讲到多变量梯度，从两…

We’re going to look at three different kinds of benchmarks, one set of c…