神经网络(Neural Networks)和深度学习(Deep Learning)是目前与人工智能(Artificial Intelligence)一起使用的两个热门流行语。人工智能世界的最新发展可以归功于这两者,因为它们在提高人工智能的智能方面发挥了重要作用。
环顾四周,你会发现周围的智能机器越来越多。由于神经网络(Neural Networks)和深度学习(Deep Learning),曾经被认为是人类强项的工作和能力现在由机器执行。今天,机器不再被制造来吃更复杂的算法,而是被喂食以发展成一个自主的、自学的系统,能够彻底改变周围的许多行业。
神经网络(Neural Networks)和深度学习(Deep Learning )在图像识别、语音识别、在数据集中寻找更深层次的关系等任务中为研究人员带来了巨大的成功。在海量数据和计算能力的帮助下,机器可以识别物体、翻译语音、训练自己识别复杂模式、学习如何制定策略和实时制定应急计划。
那么,这究竟是如何工作的呢?你知道Neutral Networks和Deep-Learning都相关吗,其实要了解Deep learning,首先要了解Neural Networks?继续阅读以了解更多信息。
什么是神经网络
神经网络基本上是一种编程模式或一组算法,使计算机能够从观察数据中学习。(Neural)神经网络类似于(Neural)人脑,通过识别模式来工作。使用机器感知、标记或聚类原始输入来解释感官数据。识别的模式是数字的,包含在向量中,图像、声音、文本等数据被转换成向量。
Think Neural Network! Think how a human brain function
如上所述,神经网络的功能就像人脑一样。它通过学习过程获得所有知识。之后,突触权重存储获得的知识。在学习过程中,网络的突触权重被改革以达到预期的目标。
就像人脑一样,神经网络(Neural Networks)像非线性并行信息处理系统一样工作,可以快速执行模式识别和感知等计算。因此,这些网络在输入/信号本质上是非线性的语音、音频和图像识别等领域表现得非常好。
简而言之,您可以将神经网络记住为能够像人脑一样存储知识并使用它进行预测的东西。(In simple words, you can remember Neural Network as something which is capable of stocking knowledge like a human brain and use it to make predictions.)
神经网络的结构
(图片来源:Mathworks)
神经网络(Networks)由三层组成,
- 输入层,
- 隐藏层,和
- 输出层。
每一层由一个或多个节点组成,如下图中的小圆圈所示。节点之间的线表示从一个节点到下一个节点的信息流。信息从输入流向输出,即从左到右(在某些情况下可能是从右到左或双向)。
输入层的节点是被动的,这意味着它们不会修改数据。他们在输入中接收单个值,并将该值复制到多个输出中。而(Whereas)隐藏层和输出层的节点是活动的。因此,他们可以修改数据。
在互连结构中,来自输入层的每个值都被复制并发送到所有隐藏节点。进入隐藏节点的值乘以权重,权重是程序中存储的一组预定数字。然后将加权输入相加以产生单个数字。神经网络可以有任意数量的层,每层可以有任意数量的节点。大多数应用程序使用最多几百个输入节点的三层结构
神经网络示例(Example of Neural Network)
考虑一个识别声纳信号中物体的神经网络,PC 中存储了 5000 个信号样本。PC 必须弄清楚这些样本是代表潜艇、鲸鱼、冰山、海石,还是什么都没有?传统的 DSP(Conventional DSP)方法将通过数学和算法来解决这个问题,例如相关性和频谱分析。
而使用神经网络时,5000 个样本将被馈送到输入层,从而导致值从输出层弹出。通过选择适当的权重,可以将输出配置为报告范围广泛的信息。例如,可能有以下输出:潜艇(是/否)、海岩(是/否)、鲸鱼(是/否)等。
使用其他权重,输出可以将物体分类为金属或非金属,生物或非生物,敌人或盟友等。没有算法,没有规则,没有程序;只有输入和输出之间的关系由所选权重的值决定。
现在,让我们了解深度学习的概念。(Now, let’s understand the concept of Deep Learning.)
什么是深度学习
深度学习基本上是神经网络(Neural Networks)的一个子集;也许您可以说一个复杂的神经网络(Neural Network),其中包含许多隐藏层。
从技术上讲,深度(Deep)学习也可以定义为一组强大的神经网络学习技术。它是指由多层、海量数据集、强大的计算机硬件组成的人工神经网络( ANN ),使复杂的训练模型成为可能。(ANN)它包含使用具有越来越丰富功能的多层人工神经网络的方法和技术类别。
深度学习网络的结构(Structure of Deep learning network)
深度(Deep)学习网络主要使用神经网络架构,因此通常被称为深度神经网络。工作“深度”的使用是指神经网络中隐藏层的数量。传统的神经网络包含三个隐藏层,而深度网络可以有多达 120-150 个。
深度(Deep) 学习(Learning)涉及为计算机系统提供大量数据,它可以用来对其他数据做出决策。这些数据是通过神经网络提供的,就像机器学习中的情况一样。深度(Deep)学习网络可以直接从数据中学习特征,而无需手动提取特征。
深度学习的例子(Examples of Deep Learning)
目前,从汽车(Automobile)、航空航天(Aerospace)和自动化(Automation)到医疗(Medical),几乎所有行业都在使用深度学习。这里有一些例子。
- 谷歌(Google)、Netflix和亚马逊(Amazon):谷歌(Google)在其语音和图像识别算法中使用它。Netflix和亚马逊(Amazon)也使用深度学习来决定你接下来想看或买什么
- 无人驾驶:研究人员正在利用深度学习网络自动检测停车标志和交通信号灯等物体。深度(Deep)学习还用于检测行人,这有助于减少事故。
- 航空航天和国防:深度学习用于识别来自卫星的目标区域,并确定部队的安全或不安全区域。
- 多亏了深度学习(Deep Learning),Facebook会自动在您的照片中找到并标记朋友。Skype 也可以实时且非常准确地翻译口语交流。
- 医学研究:医学研究人员正在使用深度学习来自动检测癌细胞
- 工业自动化(Industrial Automation):深度学习通过自动检测人或物体何时在机器的不安全距离内来帮助提高重型机械周围的工人安全。
- 电子:深度(Deep)学习正被用于自动听力和语音翻译。
阅读(Read):什么是机器学习和深度学习(Machine Learning and Deep Learning)?
结论(Conclusion)
神经网络(Neural Networks)的概念并不新鲜,研究人员在过去十年左右的时间里取得了一定的成功。但真正改变游戏规则的是深度(Deep)神经网络的发展。
通过超越传统的机器学习方法,它表明深度神经网络不仅可以由少数研究人员进行训练和试验,而且还可以被跨国科技公司采用,以便在不久的将来实现更好的创新。
Thanks to Deep Learning and Neural Network, AI is not just doing the tasks, but it has started to think!
What is Deep Learning and Neural Network
Neural Networks and Deep Learning are currently the two hot buzzwords that are being used nowadays with Artificial Intelligence. The recent developments in the world of Artificial intelligence can be attributed to these two as they have played a significant role in improving the intelligence of AI.
Look around, and you will find more and more intelligent machines around. Thanks to Neural Networks and Deep Learning, jobs and capabilities that were once considered the forte of humans are now being performed by machines. Today, Machines are no longer made to eat more complex algorithms, but instead, they are fed to develop into an autonomous, self-teaching system capable of revolutionizing many industries all around.
Neural Networks and Deep Learning have lent enormous success to the researchers in tasks such as image recognition, speech recognition, finding deeper relations in a data sets. Aided by the availability of massive amounts of data and computational power, machines can recognize objects, translate speech, train themselves to identify complex patterns, learn how to devise strategies and make contingency plans in real-time.
So, how exactly does this work? Do you know that both Neutral Networks and Deep-Learning related, in fact, to understand Deep learning, you must first understand about Neural Networks? Read on to know more.
What is a Neural Network
A Neural network is basically a programming pattern or a set of algorithms that enables a computer to learn from the observational data. A Neural network is similar to a human brain, which works by recognizing the patterns. The sensory data is interpreted using a machine perception, labeling or clustering raw input. The patterns recognized are numerical, enclosed in vectors, into which the data such are images, sound, text, etc. are translated.
Think Neural Network! Think how a human brain function
As mentioned above, a neural network functions just like a human brain; it acquires all the knowledge through a learning process. After that, synaptic weights store the acquired knowledge. During the learning process, the synaptic weights of the network are reformed to achieve the desired objective.
Just like the human brain, Neural Networks work like non-linear parallel information-processing systems which rapidly perform computations such as pattern recognition and perception. As a result, these networks perform very well in areas like speech, audio and image recognition where the inputs/signals are inherently nonlinear.
In simple words, you can remember Neural Network as something which is capable of stocking knowledge like a human brain and use it to make predictions.
Structure of Neural Networks
(Image Credit: Mathworks)
Neural Networks comprises of three layers,
- Input layer,
- Hidden layer, and
- Output layer.
Each layer consists of one or more nodes, as shown in the below diagram by small circles. The lines between the nodes indicate the flow of information from one node to the next. The information flows from the input to the output, i.e. from left to right (in some cases it may be from right to left or both ways).
The nodes of the input layer are passive, meaning they do not modify the data. They receive a single value on their input and duplicate the value to their multiple outputs. Whereas, the nodes of the hidden and output layer are active. Thus that can they modify the data.
In an interconnected structure, each value from the input layer is duplicated and sent to all of the hidden nodes. The values entering a hidden node are multiplied by weights, a set of predetermined numbers stored in the program. The weighted inputs are then added to produce a single number. Neural networks can have any number of layers, and any number of nodes per layer. Most applications use the three-layer structure with a maximum of a few hundred input nodes
Example of Neural Network
Consider a neural network recognizing objects in a sonar signal, and there are 5000 signal samples stored in the PC. The PC has to figure out if these samples represent a submarine, whale, iceberg, sea rocks, or nothing at all? Conventional DSP methods would approach this problem with mathematics and algorithms, such as correlation and frequency spectrum analysis.
While with a neural network, the 5000 samples would be fed to the input layer, resulting in values popping from the output layer. By selecting the proper weights, the output can be configured to report a wide range of information. For instance, there might be outputs for: submarine (yes/no), sea rock (yes/no), whale (yes/no), etc.
With other weights, the outputs can classify the objects as metal or non-metal, biological or non-biological, enemy or ally, etc. No algorithms, no rules, no procedures; only a relationship between the input and output dictated by the values of the weights selected.
Now, let’s understand the concept of Deep Learning.
What is a Deep Learning
Deep learning is basically a subset of Neural Networks; perhaps you can say a complex Neural Network with many hidden layers in it.
Technically speaking, Deep learning can also be defined as a powerful set of techniques for learning in neural networks. It refers to artificial neural networks (ANN) that are composed of many layers, massive data sets, powerful computer hardware to make complicated training model possible. It contains the class of methods and techniques that employ artificial neural networks with multiple layers of increasingly richer functionality.
Structure of Deep learning network
Deep learning networks mostly use neural network architectures and hence are often referred to as deep neural networks. Use of work “deep” refers to the number of hidden layers in the neural network. A conventional neural network contains three hidden layers, while deep networks can have as many as 120- 150.
Deep Learning involves feeding a computer system a lot of data, which it can use to make decisions about other data. This data is fed through neural networks, as is the case in machine learning. Deep learning networks can learn features directly from the data without the need for manual feature extraction.
Examples of Deep Learning
Deep learning is currently being utilized in almost every industry starting from Automobile, Aerospace, and Automation to Medical. Here are some of the examples.
- Google, Netflix, and Amazon: Google uses it in its voice and image recognition algorithms. Netflix and Amazon also use deep learning to decide what you want to watch or buy next
- Driving without a driver: Researchers are utilizing deep learning networks to automatically detect objects such as stop signs and traffic lights. Deep learning is also used to detect pedestrians, which helps decrease accidents.
- Aerospace and Defense: Deep learning is used to identify objects from satellites that locate areas of interest, and identify safe or unsafe zones for troops.
- Thanks to Deep Learning, Facebook automatically finds and tags friends in your photos. Skype can translate spoken communications in real-time and pretty accurately too.
- Medical Research: Medical researchers are using deep learning to automatically detect cancer cells
- Industrial Automation: Deep learning is helping to improve worker safety around heavy machinery by automatically detecting when people or objects are within an unsafe distance of machines.
- Electronics: Deep learning is being used in automated hearing and speech translation.
Read: What is Machine Learning and Deep Learning?
Conclusion
The concept of Neural Networks is not new, and researchers have met with moderate success in the last decade or so. But the real game-changer has been the evolution of Deep neural networks.
By out-performing the traditional machine learning approaches it has showcased that deep neural networks can be trained and trialed not just by few researchers, but it has the scope to be adopted by multinational technology companies to come with better innovations in the near future.
Thanks to Deep Learning and Neural Network, AI is not just doing the tasks, but it has started to think!