– you really need to start using neural networks. When the patterns get really complex,

你真的需要開始使用神經網絡。如果模式真的很複雜，

neural nets start to outperform all of their competition. Plus, GPUs can train them faster

神經網絡開始會勝過他們所有的競爭者。再者，GPUs 可以訓練他們比以前更快。

than ever before! Let's take a look.

讓我們來看一下吧！

Neural nets truly have the potential to revolutionize the field of Artificial Intelligence. We all

神經網絡真的很有潛力可以去大力改革人工智慧的領域。我們都

know that computers are very good with repetitive calculations and detailed instructions, but

知道，電腦對於重複計算跟精細的結構非常在行，但是

they've historically been bad at recognizing patterns. Thanks to deep learning, this is

自古以來他們在辨識模式上都很糟。感謝深度學習，這些

all about to change.

都將改變

If you only need to analyze simple patterns, a basic classification tool like an SVM or

如果你只需要去分析簡單的模式，一個基本的分類工具像是 SVM 或是

Logistic Regression is typically good enough. But when your data has 10s of different inputs

邏輯回歸就很夠了。但當你的資料有十個以上不同的輸入(指令)

or more, neural nets start to win out over the other methods. Still, as the patterns

甚或更多時，神經網絡開始贏過其他的方法。並且，雖然模式

get even more complex, neural networks with a small number of layers can become unusable.

變得更為複雜，神經網絡用少少的幾層結構就會無法勝任了。

The reason is that the number of nodes required in each layer grows exponentially with the

原因是因為節點的數量在每一階層中，都會隨著資料中可能模式的數量而呈指數的成長。

number of possible patterns in the data. Eventually training becomes way too expensive and the

最終訓練會變成太過昂貴的方式，並且

accuracy starts to suffer. So for an intricate pattern – like an image of a human face,

準確性開始出現缺失。所以為了一個複雜的模式－像是一個人臉的圖像，

for example – basic classification engines and shallow neural nets simply aren’t good

舉例－基礎的分類引擎以及淺層的神經網絡已經完全不夠好了

enough – the only practical choice is a deep net.

－唯一實用的選擇是是深度網路。

Have you ever run into a wall when trying to work with highly complex data? Please comment

當你試圖要跟複雜度更高的資料工作時，是否有撞牆的感覺？請留下評論

and let me know your thoughts.

並讓我知道你的想法。

But what enables a deep net to recognize these complex patterns? The key is that deep nets

但究竟是什麼讓深度的網路可以去辨識那些複雜的模組呢？關鍵就是深度網路

are able to break the complex patterns down into a series of simpler patterns. For example,

能把複雜的網路破解成一系列的簡化模式。例如，

let's say that a net had to decide whether or not an image contained a human face. A

讓我們來說說一個網路必須要決定在一張圖片中是否包含一個人臉。一個

deep net would first use edges to detect different parts of the face – the lips, nose, eyes,

深的網路首先會使用銳化去發現臉的不同部位－嘴唇、鼻子、眼睛、

ears, and so on – and would then combine the results together to form the whole face.

耳朵之類的－然後再將這些結果組合成一個完整的臉。

This important feature – using simpler patterns as building blocks to detect complex patterns

這重要特徵－利用簡化模式去建構區塊以偵查出複雜的模組

– is what gives deep nets their strength. The accuracy of these nets has become very

－這就是讓深度網絡有利的原因。此網絡的精確度就變得很是

impressive – in fact, a deep net from google recently beat a human at a pattern recognition

令人印象深刻－事實上，Google 的網絡最近在模式辨識競賽中擊敗了一個人類

challenge.

It’s not surprising that deep nets were inspired by the structure of our own human

深層網絡的靈感來自於我們自己的人類大腦結構，這並不奇怪

brains. Even in the early days of neural networks, researches wanted to link a large number of

即使是在神經網絡的早期，也有一些研究希望通過分層網絡將大量

perceptrons together in a layered web – an idea which helped improve their accuracy.

感知器連接在一起，－這是一種有助於提高其精度的想法。

It is believed that our brains have a very deep architecture and that we decipher patterns

研究相信我們的大腦有一個非常深度的結構，且我們在解譯模組時

just like a deep net – we detect complex patterns by first detecting, and combining,

就像是一個深度網絡－我們藉由先辨識跟結合簡單的部份來辨識複雜著模組。

the simple ones.

There is one downside to all of this – deep nets take much longer to train. The good news

所但這一切都有一個缺點 - 深網需要更長的時間訓練。好消息

is that recent advances in computing have really reduced the amount of time it takes

是在計算上的進展已經減少了很大量的時間，需要

to properly train a net. High performance GPUs can finish training a complex net in

適當的去訓練一個網絡。高性能 GPU 可以完成一個複雜網絡的訓練在

under a week, when fast CPUs may have taken weeks or even months.

一週之內，當快速的 CPU 可能還需要幾週甚至幾個月時

Before we talk more about the various Deep Learning models, we're going to briefly

在我們討論更多各種深度學習模式之前，我們要簡短地

discuss which types of deep nets are suitable for different machine learning tasks. That's

談論怎樣的深度網路對不同的機器學習任務是適合的。

coming up in the next video.

這將會在下支影片中出現