most practitioners use a procedure called stochastic gradient descent (SGD).

, The chain rule of derivatives tells us how two small effects (that of a small change of x on y, and that of y on z) are composed.

The backpropagation procedure to compute the gradient of an objective function with respect to the weights of a multilayer stack of modules is nothing more than a practical application of the chain rule for derivatives.

根据评论区 @山丹丹@啸王 的提醒，更正了一些错误（用斜体显示），在此谢谢各位。并根据自己最近的理解，增添了一些东西（用斜体显示）。如果还有错误，欢迎大家指正。第一个问题：为什么引入非线性激励函数？如果不用激励函数（其实相当于激励函数是f(x) = x），在这种情况下你每一层输出都是上层输入的线性函数，很容易验证，无论你神经网络有多少层，输出都是输入的线性组合，与没有隐藏层效果相当，这种情况就是最原始的感知机（Perceptron）了。正因为上面的原因，我们决定引入非线性函数作为激励函数，这样深层神经网络就有意义了（不再是输入的线性组合，可以逼近任意函数）。最早的想法是sigmoid函数或者tanh函数，输出有界，很容易充当下一层输入（以及一些人的生物解释balabala）。第二个问题：为什么引入Relu呢？第一，采用sigmoid等函数，算激活函数时（指数运算），计算量大，反向传播求误差梯度时，求导涉及除法，计算量相对大，而采用Relu激活函数，整个过程的计算量节省很多。第二，对于深层网络，sigmoid函数反向传播时，很容易就会出现梯度消失的情况（在sigmoid接近饱和区时，变换太缓慢，导数趋于0，这种情况会造成信息丢失，参见 @Haofeng Li 答案的第三点），从而无法完成深层网络的训练。第三，Relu会使一部分神经元的输出为0，这样就造成了网络的稀疏性，并且减少了参数的相互依存关系，缓解了过拟合问题的发生（以及一些人的生物解释balabala）。当然现在也有一些对relu的改进，比如prelu，random relu等，在不同的数据集上会有一些训练速度上或者准确率上的改进，具体的大家可以找相关的paper看。多加一句，现在主流的做法，会在做完relu之后，加一步batch normalization，尽可能保证每一层网络的输入具有相同的分布[1]。而最新的paper[2]，他们在加入bypass connection之后，发现改变batch normalization的位置会有更好的效果。大家有兴趣可以看下。

Unsupervised Learning of 3D Structure from Images Authors: Danilo Jimenez Rezende, S. M. Ali Eslami, Shakir Mohamed, Peter Battaglia, Max Jaderberg, Nicolas Heess (Submitted on 3 Jul 2016) Abstract: A key goal of computer vision is to recover the underlying 3D structure from 2D observations of the world. In this paper we learn strong deep generative models of 3D structures, and recover these structures from 3D and 2D images via probabilistic inference. We demonstrate high-quality samples and report log-likelihoods on several datasets, including ShapeNet [2], and establish the first benchmarks in the literature. We also show how these models and their inference networks can be trained end-to-end from 2D images. This demonstrates for the first time the feasibility of learning to infer 3D representations of the world in a purely unsupervised manner.

When building a unified vision system or gradually adding new capabilities to a system, the usual assumption is that training data for all tasks is always available. However, as the number of tasks grows, storing and retraining on such data becomes infeasible. A new problem arises where we add new capabilities to a Convolutional Neural Network (CNN), but the training data for its existing capabilities are unavailable. We propose our Learning without Forgetting method, which uses only new task data to train the network while preserving the original capabilities. Our method performs favorably compared to commonly used feature extraction and fine-tuning adaption techniques and performs similarly to multitask learning that uses original task data we assume unavailable. A more surprising observation is that Learning without Forgetting may be able to replace fine-tuning as standard practice for improved new task performance.

Low-shot visual object recognition

OpenAI is a non-profit artificial intelligence research company. Our goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.

let you share a URL that navigates directly to the selection you shared.

¿Podría contar alguna anécdota de su participación en el desarrollo de Deep Blue? Hay muchas anécdotas, pero son largas de contar. Solo te diré que el día que empecé a trabajar para IBM para enumerar los puntos débiles de Kaspárov el folio se quedó en blanco.