ADM-201 dump PMP dumps pdf SSCP exam materials CBAP exam sample questions

USB神经网络 – 译学馆
未登录,请登录后再发表信息
最新评论 (0)
播放视频

USB神经网络

Neural Network Prototyping On the Go

您正在收看的是冷流科技
You are watching ColdFusion TV.
冷流科技
[Music playing]
大家好 欢迎再次收看ColdFusion视频栏目
Hi! Welcome to another ColdFusion video.
今天的节目内容有一点点不同
Today’s episode is going to be something a little bit different.
英特尔公司找到我
Intel approached me
想让我谈谈神经网络的
and asked me to talk about the concept
概念及应用方式
and application of neural networks
以便和他们的英特尔Movidius神经计算棒结合
to tie in with their Intel Movidius Neural Compute Stick.
首来
First,
看看这个设备到底是什么
we’re going to cover what the device actually is
之后我们会深入了解神经网络
and then we’re gonna dig a bit deeper into neural networks
并挖掘更深的用途
and what they’re use for.
那我们开始吧
So let’s begin.
那么
So
什么是英特尔Movidius神经计算棒呢
what is the Intel Movidius Neural Compute Stick?
它是一种紧凑型设备
It’s a compact device
比标准USB驱动器大不了多少
not much bigger than a standard USB drive
连接电脑
that plugs into a computer
使用usb3.0接口
through USB 3 port.
该设备支持快速原型验证
It enables rapid prototyping validation
以及深度神经网络界面应用程序的开发
and development of deep neural network interface applications
而且不需要用户连接到云端
and it doesn’t require users to connect to the cloud.
该设备基本上是一个
It’s basically hardware designed
高效运行某些特定类型的人工智能系统的硬件
to run certain types of AI systems super efficiently.
好 接着我们花一分钟
OK, so let’s back up for a minute
回顾一下到底什么是神经网络
and explored a neural network actually is.
简而言之
In simple terms,
神经网络是一种算法
a neural network is an algorithm
它基于机器学习的原理
based on the principles of machine learning.
那么 机器学习又是什么?
So what is machine learning?
就是计算机程序或人工智能
It’s the ability for a computer program or AI
通过经验学习
to gradually learn by experience
逐渐完成程序员设定的目标的能力
towards a goal established by the human programmer.
例如
For instance,
如果你想训练人工智能辨别猫和狗
if you wanted to train an AI to tell the difference between a cat and a dog,
首先要建立基础参数
you start with the basic parameters.
然后逐渐完善并测试
Then you have it built, test it,
再选出最佳模型
then take the top-performing models, rebuild
并进行改造和微调
and fine-tune from there.
不断重复这个过程
This process happened over and over again,
最终生成的就是我们所认为的
ultimately generating what we consider
经历成千上万次重复后的更智能的算法
more intelligent algorithms after thousands of iterations.
这个过程产生了人工神经元
This process generates artificial neurons,
也被称为节点
otherwise known as nodes, .
就是连接数据的点
which are connection points of data
每个节点都带有一个方程式
Each of these nodes consists of an equation
并含有输入端及输出端
and have an input and an output.
节点排列成图层
The nodes are arranged in layers
图层和节点共同构成神经网络
that all together make up the neural network.
神经网络最迷人的地方在于
What’s really fascinating about neural networks is that
网络中有很多的连接 而且及其精确
is that there’s so many of these connections and they’re so elaborate
精确到原始程序员都不能完全确定
that the original programmers aren’t entirely sure about
这些连接是如何产生的或者作用是什么
how they come about or what they even mean.
最终的结果就是算法能像人一样学习
The end result is an algorithm that can learn things
识别图案
recognize patterns
以及做出决定
and make decisions in a human way
而且不一定需要编程就可以精确学习
and it doesn’t have to be programmed to learn explicitly.
它可以自主学习
It learns by itself
就像大脑一样
just like a brain.
自2012年以后 其发展更为惊人
And even more staggering since 2012,
神经网络的复杂性
the general complexity of neural networks .
普遍增强了500倍
has increased 500 times
神经网络建立起来后
When a neural network is formed,
可以完成高度专门化的任务
it can do highly specialized tasks,
比如我们之前提到的识别猫和狗这样的简单例子
such as the previous simple example of a cat vs. dog sorter.
该程序被称为视觉处理
This is an application known as visual processing.
生成一个神经网络的过程
The development of generating a neural network
被称为需要大量数据的训练阶段
is known as the training phase and involves vast amounts of data.
训练完成后
After this phase,
神经网络需要在真实的app中进行测试
the neural network needs to be tested in real-world applications
来观察它如何实现实时响应
to see how it responds in real-time
并判断它的效果是否理想
and if it works as desired.
一般来说
Traditionally,
你需要连接到云端进行这个过程
you’d need to connect to a cloud service to do this.
但如果有别的方式呢?
But what if there’s another way?
这就是英特尔Movidiua神经计算棒产生的原因
This is where the Intel Movidius Neural Compute Stick comes in.
这个设备的妙用在于
What’s fascinating about this device is that
你可以通过它获得深度神经网络的功能
you can gain on device deep neural network capabilities
却不用耗费太多能量
without using much power.
事实上正如我之前所说
In fact as I mentioned before,
它被认为是一种低功率设备 这种设备非常酷
this is considered a low power device, which is pretty cool
因为在大多情境下 你需要利用大量的服务器群进行计算
because in most scenarios you’d be utilizing massive server farms.
英特尔神经计算棒自带一个专门的硬件视觉处理装置
Intel’s Neural Compute Stick has a specialized onboard hardware visual processing unit
性能超过100吉帕
capable of more than 100 gigaflops of performance
嵌在一个1瓦特的外壳中
within a 1 watt power envelope.
英特尔Modivius神经计算棒可以
The Intel Movidius Neural Compute Stick can
计算神经网络所需的大量数据
calculate the vast amounts of data required in a neural network
而不一定要连接到云端甚至联网
without having to connect to the cloud or even be online.
这相当于增强了保密性和可靠性
This equates to an increase in privacy and reliability.
这意味着无论你在办公室工作还是出差
And it means whatever you’re working on stays local and mobile,
工作地点固定带来的变量更少
there are fewer variables inherent with working local
同时使工作更顺畅
and also keeps things running smoother,
因此你就可以更快解决问题
so that you can troubleshoot quicker,
因为你立刻就能消除更大的隐患
since you eliminate some of the bigger potential factors right away.
所以英特尔神经计算棒
So Intel’s Neural Compute Stick
并不是取代传统的神经网络
doesn’t replace the traditional style of the neural network there.
神经计算棒的作用在于填补了更基础的需要
What it does is fill a need on a more basic level.
从根本上缩小了昂贵的庞大网络与
Essentially closing the gap between massive and expensive networks
更多实验项目之间的差距
and more experimental projects.
例如 我和我朋友在运行一个小神经网络
For example, me and my friend had a little go at running a neural network
该网络能把手绘艺术风格转化为照片
that allows hand-painted art styles to be transferred to photos.
最后 英特尔神经计算棒低功率的本质在于
Lastly, the low-power nature of the Intel Neural Compute Stick
它提供了更多的可能性
opens up even more doors.
是的 这可以节约能源 但更重要的是
Yes, it conserves energy, but more important than that,
这意味着消费者的紧凑型设备
it means compact consumer devices
可以受益于采用新型人工智能供能的应用程序
can benefit from new types of AI powered applications.
虚拟现实 混合现实
Areas like virtual and mixed reality,
无人机 机器人及安全系统这些领域
drones, robotics and security systems,
基本是在神经网络上的不断竞争
basically it’s neural network competing on the go.
毕竟发展中的事物才是强大的
At all likeliest is powerful,
结合英特尔神经计算棒就更是如此了
and it becomes more so when paired with Intel’s Neural Compute SDK.
所以如果你想深入地了解英特尔人工智能
So, if you’d like to know more about Intel AI,
我会在下面的说明中放几个链接
I’ve included some links in the description below.
总之 以上就是视频的全部内容了
Anyway, that’s all for this video.
如果你知道有人对这方面感兴趣的话
If you know someone who would be interested in this,
自由将视频安利给他们吧
feel free to pass it along to them.
在下方评论留言 告诉我你对神经网络的想法
Leave a comment in the description and tell me what you think about neural networks.
这里是Dagogo 你观看的是冷流科技
This has been Dagogo and you’ve been watching ColdFusion.
下期节目我们很快会再见的
I’ll see you again soon for the next video.
谢谢大家 天天开心!
Cheers guys. Have a good one!
冷流科技 新的想法
ColdFusion. It’s a new thinking.

发表评论

译制信息
视频概述

该视频主要介绍了英特尔推出的名为神经计算棒的新型设备。该设备可以和神经网络系统连接,从而大幅降低计算所需的功耗

听录译者

在濑

翻译译者

沐子

审核员

审核员 LD

视频来源

https://www.youtube.com/watch?v=FDguqBh3V6g

相关推荐