One of the greatest road-blocks for newcomers to data science and deep learning is the multitude of terms present on even the most introductory courses. The field has a steep learning curve, as most concepts are intertwined and complementary. To make things worse, most newcomers have little to no machine learning background, making the task even harder. In this article, I break down some of field’s most important concepts in a sequential manner, linking each new term to the last.

最大的路障为初来乍到的数据科学和深度学习的ØNE存在，即使是最入门课程方面的群众。 该领域具有陡峭的学习曲线，因为大多数概念是相互交织和互补的。 更糟的是，大多数新手几乎没有机器学习背景，这使任务变得更加艰巨。 在本文中，我将按顺序分解一些领域中最重要的概念，将每个新术语与最后一个术语联系起来。

Let’s start with the definition of…

让我们从…的定义开始

Intelligence: So far, no commonly accepted definition for intelligence exists. The definition is an on-going debate that spans a wide range of fields and disciplines, such as philosophy, psychology, sociology, and computer science. However, although no definition is widely accepted, science has been trying to mimic intelligence for decades, creating…

情报：到目前为止，还没有公认的情报定义 。 该定义是一个持续不断的辩论，涵盖了哲学，心理学，社会学和计算机科学等众多领域和学科。 然而，尽管没有一个定义被广泛接受，但数十年来，科学一直在尝试模仿智力，从而创造出……

Artificial Intelligence (AI): The field within Computer Science that studies “intelligent behavior” in machines. In other words, systems that behave in non-trivially predictable ways. Systems that react and (appear to) reason about themselves and the world around them. This definition is vague as we can’t even define intelligence properly. However, something that we can define is learning, which brings us to…

人工智能(AI)：计算机科学领域，研究机器中的“智能行为”。 换句话说，以非平常可预测的方式运行的系统。 对自己和周围世界做出React并(表现出)理性的系统。 这个定义含糊不清，因为我们甚至无法正确定义智能。 但是，我们可以定义的是学习，这使我们能够…

Machine Learning (ML): The task of showing the inputs and outputs of a problem to an algorithm and letting it learn how to solve it. For instance, to create an algorithm that recognizes dogs and cats in pictures by showing it hundreds of photos of pets. Within, ML, a growing approach is:

机器学习(ML)：将问题的输入和输出显示给算法并让其学习如何解决的任务。 例如，创建一种算法，通过显示数百张宠物照片来识别图片中的猫狗。 在ML中，一种不断发展的方法是：

Deep Learning (DL): The idea of stacking multiple learning algorithms to jointly solve a difficult task. When we are kids, we learn the alphabet, then we learn to read simple words, then full sentences, etc. With practice, we are reading books and forming opinions. In between letters and fluency, several learning steps are involved. In imitation, Deep learning algorithms approach learning by breaking the task down into sequential problems, each one building an increased level of abstraction.

深度学习(DL)：堆叠多种学习算法以共同解决一项艰巨任务的想法。 小时候，我们学习字母，然后学习阅读简单的单词，然后阅读完整的句子，等等。通过练习，我们正在阅读书籍并形成意见。 在字母和流利度之间，涉及几个学习步骤。 在模仿中，深度学习算法通过将任务分解为顺序的问题来进行学习，每个问题都建立了更高的抽象水平。

In practice, DL and ML algorithms are frequently the tools of choice for…

在实践中，DL和ML算法通常是...的选择工具

Data Science (DS): The study of data, such as acquisition, cleaning, storing, viewing, and processing data. A common misconception is that DL and DS are the same things, as many DS courses feature DL prominently. ML and DL are tools for the data scientist, as is SQL, NumPy and, etc.

数据科学(DS)：对数据的研究，例如采集，清理，存储，查看和处理数据。 一个常见的误解是DL和DS是同一件事，因为许多DS课程都突出显示DL。 ML和DL是数据科学家的工具 ，SQL，NumPy等也是如此。

Data Engineering, Data Analytics, Data Analysis, Cognitive Computing: This area is full of lesser-known synonyms, and confusing terms. Some seem to only exist to sound new and fresh, some make reference to specific tasks within DS. For instance, “analytics” usually refers to exploratory analysis while “engineering” refers to implementing systems for production use. Stick to the widely used terms whenever in doubt, such as “data science”.

数据工程，数据分析，数据分析，认知计算：该领域充满了鲜为人知的同义词和令人困惑的术语。 有些似乎只是听起来新鲜而新，有些则引用了DS中的特定任务。 例如，“分析”通常是指探索性分析，而“工程”是指针对生产使用的实施系统。 如有疑问，请坚持使用广泛使用的术语，例如“数据科学”。

Artificial Neural Network (NN): Among several ways of implementing deep learning, neural networks are by far the most popular. In short, they are a stack of simple learning algorithms (called layers) that sequentially process the input, producing an output. This embodies the idea of deep learning by design, as each layer learns a more refined understanding of the input.

人工神经网络(NN)：在实现深度学习的几种方式中，神经网络是迄今为止最受欢迎的方法。 简而言之，它们是一堆简单的学习算法(称为“层”)，它们依次处理输入并产生输出。 这体现了通过设计进行深度学习的思想，因为每一层都可以更好地理解输入。

Network Layers: All neural networks are made up of layers (as shown in the image above). A layer defines an operation that takes some inputs, some parameters, and produces a set of outputs. For instance, the…

网络层：所有神经网络均由层组成(如上图所示)。 一层定义了一个操作，该操作接受一些输入，一些参数并产生一组输出。 例如...

Dense Layer: Is the layer that receives a vector (input) and multiplies it by a matrix (parameters), producing another vector (outputs). While this is one of the most widely used layers in practice, it has a problem, it is:

密集层：是接收向量(输入)并将其乘以矩阵(参数)以产生另一个向量(输出)的层。 尽管这是实践中使用最广泛的层之一，但它有一个问题，那就是：

Linear: Something is linear when it can be decomposed as independent parts, this makes them easy to study. However, a linear system is no more interesting than the parts that make it. “A+B” is no more interesting than “A” and “B” themselves. This is opposed by:

线性 ：可以分解为独立部分的事物是线性的，这使它们易于学习。 但是， 线性系统并不比构成线性系统的零件有趣 。 “ A + B”并不比“ A”和“ B”本身有趣。 与之相反的是：

Non-Linear: A system is non-linear when its parts are intertwined as a complex whole. A non-linear system cannot be easily factored. Hence, the whole is more than the sum of its parts. For instance, “sine(A + B)” cannot be broken down into “A” terms and “B” terms. Thankfully, any linear system can be made non-linear by means of an…

非线性：当系统的各个部分交织成一个复杂的整体时，它就是非线性的。 非线性系统不容易分解。 因此， 整体不仅仅是部分的总和 。 例如，“正弦(A + B)”不能细分为“ A”项和“ B”项。 幸运的是，任何线性系统都可以通过...

Activation Function: These are ordinary non-linear functions used in DL that have the sole purpose of receiving input and making it non-linear. In the above example, we made “A + B” non-linear by using the sine function. Within DL, the most famous activation function is the:

激活函数：这些是DL中使用的普通非线性函数，其唯一目的是接收输入并使之成为非线性。 在上面的示例中，我们使用正弦函数使“ A + B”成为非线性。 在DL中，最著名的激活功能是：

ReLU: Short of Rectified Linear Unity, defined as ReLU(x) = max(0, x). This is one of the simplest (and most efficient) ways of making something non-linear. That’s why it is so popular: it is simple and fast. Combining dense layers and ReLU activations, we can build a…

ReLU：缺少线性校正单位，定义为ReLU(x)= max(0，x)。 这是使非线性事物最简单(也是最有效)的方法之一。 这就是为什么它如此流行：它既简单又快速。 结合密集层和ReLU激活，我们可以构建一个...

Dense Network / Fully-Connected Network / Multi-Layer Perceptron: These are all synonyms for the basic neural network: a collection of dense layers interleaved with activation functions. It is important to highlight that, without activation functions, a network would be no more interesting than a single dense layer. The non-linearities are the glue that creates a powerful model out of ordinary parts.

密集网络/全连接网络/多层感知器：这些都是基本神经网络的同义词：密集层与激活函数交错的集合。 重要的是要强调一点，没有激活功能，网络将不会比单个密集层有趣。 非线性是在普通零件之外创建强大模型的粘合剂。

Out of the synonyms for neural networks, one has historical significance…

在神经网络的同义词中，具有历史意义……

Perceptron Model: In 1958, Frank Rosenblatt created a computational model of a neuron: the perceptron, which is the basis for most of what came afterward. Simply put, the perceptron is a weighted sum followed by an activation function. Formally, P(x; w) = a(x ⋅ w). This means the perceptron receives an “x” vector, has a set of weights “w”, and computes the dot product of “x” with “w” (the weighted sum) and feeds it to “a”, the activation function.

感知器模型： 1958年， 弗兰克·罗森布拉特 ( Frank Rosenblatt)创建了神经元的计算模型：感知器，这是后来出现的大多数事情的基础。 简而言之，感知器是一个加权和，后跟一个激活函数。 形式上，P(x; w)= a(x⋅w)。 这意味着感知器接收 “ x”向量， 具有一组权重“ w”，并计算 “ x”与“ w”(加权和)的点积，并将其馈送到激活函数“ a”。

Dense Layer (revisited): If an input is fed to many perceptrons at once, as in the figure above, we can simplify the math by “joining” the weights of each perceptron as a matrix, creating the dense layer: D(x; W) = a(x ⋅ W). Using this definition, we can build a two-layers network by feeding the result of a dense layer into another, such as: D(D(x;W₁);W₂) = a(a(x ⋅ W₁)⋅W₂). This showcases an important notion: neural networks are just math functions.

密集层(重新访问)：如上图所示，如果一次将输入馈送给许多感知器，我们可以通过将每个感知器的权重作为矩阵“连接”在一起，创建密集层：D(x; W)= a(x⋅W)。 使用这个定义，我们可以通过将一个致密层的结果馈入另一个网络来构建一个两层网络，例如：D(D(x;W₁); W 2)= a(a(x·W₁)·W 2)。 这表明了一个重要的概念： 神经网络只是数学函数。

In this expression, what we haven’t given attention to yet are the…

在此表达式中，我们尚未关注的是……

Weights: Machine learning models define an “operation” that is guided by a set of “weights”. By changing a model’s weights, we are able to make it do different things. For instance, we might train the same network to recognize cats and dogs or birds and fishes. What changes are the weights, not the network. This brings us to…

权重：机器学习模型定义了以一组“权重”为指导的“操作”。 通过更改模型的权重，我们可以使其执行不同的操作。 例如，我们可能会训练同一网络来识别猫，狗或鸟和鱼。 变化的是权重，而不是网络。 这将我们带到…

Training: The actual “learning” is performed by the training loop. In simplified terms, training means to feed inputs to a model, collect its outputs, compare them with the expected outputs, and change the weights to make to correct the outputs. The main component of this process is the…

培训：实际的“学习”是由培训循环执行的。 简而言之，训练意味着将输入提供给模型，收集其输出，将其与预期输出进行比较，并更改权重以更正输出。 该过程的主要组成部分是……

Loss Function: The loss, or error, is a function that measures the “wrongness” of the model. In practice, all we want is to reduce the loss of our models. In other words, we want them to be as correct as possible. To accomplish that, we use a mathematical tool known as…

损失函数：损失或错误是一种度量模型“错误”的函数。 实际上，我们想要的是减少模型的损失。 换句话说，我们希望它们尽可能正确。 为此，我们使用一种称为…的数学工具。

Gradient Descent: Given a model “M”, its weights “θ”, the inputs “x”, the expected outputs “y” and a loss function “L”, we can optimize the weights to reduce the loss function by considering the gradient of “L(y, M(x; θ))”, known as “∇L”. This “∇L” indicates how the loss changes as θ changes. Hence, to reduce the loss we subtract ∇L from θ. This process creates an “improved θ”. If we repeat this process over and over, we get…

梯度下降：给定模型“ M”，权重“θ”，输入“ x”，预期输出“ y”和损失函数“ L”，我们可以通过考虑梯度来优化权重以减少损失函数“ L(y，M(x;θ))”中的“ L”。 该“∇L”表示损耗如何随着θ变化而变化。 因此，为了减少损耗，我们从θ中减去∇L。 该过程产生“改善的θ”。 如果我们一遍又一遍地重复此过程，则会得到……

Epochs: Each time we train our model with every data example we have, we complete an “epoch”. Models are typically trained for tens to hundreds of epochs until their losses are reduced to tiny values.

纪元：每次我们使用每个数据示例训练模型时，都会完成一个“纪元”。 模型通常经过数十到数百个时期的训练，直到它们的损失减少到极小的值。

Backpropagation: The above procedure is quite complicated if you are not familiar with the math, yet, I presented a simplified version of it. The complete version is known as backpropagation and has the added complexity that each layer has its own gradient (instead of a single gradient for the whole model, as presented). In this context, some good terms to know are:

反向传播：如果您不熟悉数学，上述过程非常复杂，但是，我提供了它的简化版本。 完整版本称为反向传播，并具有额外的复杂性，即每个层都有自己的渐变(而不是整个模型的单个渐变，如图所示)。 在这种情况下，要知道的一些好术语是：

SGD / RMSprop / Adam / Nadam / Radam / Ranger / etc: These are optimizers: algorithms that enhance backpropagation to make it faster and better. Understanding these methods is an advanced topic. In practice, everyone uses Adam 95% of the time. Another useful term to know is…

SGD / RMSprop /亚当/纳丹/拉丹/游侠/等 ：这些是优化程序 ：增强反向传播以使其更快更好的算法。 了解这些方法是高级主题。 实际上，每个人都有95％的时间使用Adam。 另一个有用的名词是……

Learning Rate: During training, it works best to update weights in baby steps. While this seems counter-intuitive, it improves convergence. Common values are 0.01, 0.001 and 0.0001. The learning rate is often a parameter of the optimizer. Another common practice during training is to divide it in…

学习率：在训练过程中，最好以婴儿步长更新体重。 尽管这似乎违反直觉，但它改善了收敛性。 常用值为0.01、0.001和0.0001。 学习率通常是优化器的参数。 培训期间的另一种常见做法是将其分为…

Batches: We often have more data than we can fit in memory or in GPU. Thus, we cannot compute the gradient for all examples we have. Instead, we can compute it for a small subset (a batch) and apply it. For instance, a thousand examples can be divided into sixteen batches of 64 elements, yielding sixteen training gradients.

批处理：我们的数据经常超出内存或GPU的容量。 因此，我们无法为所有示例计算梯度。 取而代之的是，我们可以为一个小的子集(一个batch )计算它并应用它。 例如，可以将一千个示例划分为16个批次的64个元素，从而产生16个训练梯度。

That’s all for now. If you are serious about learning deep learning, many other terms were left out of this (brief) summary. In particular, I did not venture into the magical world of convolutional neural networks, which has many many names, such as convolutions, poolings, blocks, and several widely popular architectures, such as ResNets, MobileNets, InceptionNets and so on.

牛逼帽子就是现在。 如果您认真学习深度学习，那么此(简要)摘要中不包含许多其他术语。 特别是，我并没有冒险进入卷积神经网络的神奇世界，卷积神经网络有很多名称，例如卷积，池，块和几种广泛流行的体系结构，例如ResNets，MobileNets，InceptionNets等。

Several of these terms also reveal an entire world of possibilities. While I discussed what a loss function is, we didn’t even scratch the surface of how many of them exist. The same goes for activation functions.

这些术语中的几个还揭示了整个可能性。 当我讨论损失函数是什么时，我们甚至没有刮过其中存在多少个函数的表面。 激活功能也是如此。

As mentioned in the beginning, this article is meant as an introduction, covering most of the fundamental aspects in which everything else is built upon. I hope this has been an interesting read for you as much as it as interesting for me to write.

如开头所述，本文旨在作为介绍，涵盖构建其他所有内容的大部分基本方面。 我希望这对您来说既有趣，对我来说也很有趣。

Stay safe and subscribe for more deep learning resources

保持安全并订阅更多深度学习资源

Thank you o/

谢谢o /