The dataset

دوره: یادگیری عمیق با TensorFlow / فصل: The MNIST example / درس 1

The dataset

توضیح مختصر

  • زمان مطالعه 0 دقیقه
  • سطح خیلی سخت

دانلود اپلیکیشن «زوم»

این درس را می‌توانید به بهترین شکل و با امکانات عالی در اپلیکیشن «زوم» بخوانید

دانلود اپلیکیشن «زوم»

فایل ویدیویی

متن انگلیسی درس

I know it has been a long way before we came back to coding but all the lessons so far were necessary

to understand what we are doing.

Your patients will now pay off.

When we started modeling it would have been easy to introduce tensor flow provide you a template and

tell you put this here.

Put that there.

However now that you know the theory you can truly understand the whole process and everything going

on.

All right.

The problem we have chosen is the hello world of machine learning.

As often this is one of the first problems students encounter.

It is called M classification the amnesty dataset consists of around 70000 images of handwritten digits

since we have 10 digits.

There are 10 classes from 0 to 9.

Our objective is to build an algorithm that takes as input an image and then correctly determines which

number is shown in that image.

There are a few reasons we started with this algorithm initially.

We wanted to create an example not used before but then we thought there is a reason.

Amnesty is the hello world of machine learning.

First it is a visual problem.

You can see the data and you know what to expect.

This makes the problem easy to define and understand.

Second it has been tested by almost everyone who ever touched a deep learning algorithm.

We want our students to know what people are talking about when they say you know when I first did the

amnesty I thought this and that third.

It is easy for you to build up to convolution or neural networks from the amnesty example and finally

the dataset is both sufficiently large and clean.

The latter meaning that there are no missing values wrong labels smudged pictures etc.

Of course we

should credit the creators.

The dataset was developed by Yann Le coon Carina Cortez and Christopher Burgess.

You can find more about it on Yann Le Koons website at yen dot lacuna dot com.

Yan is one of the founding fathers of convolution on neural networks and modern image recognition.

He is the director of A.I.

research at Facebook.

As you can imagine doing machine learning at Facebook involves a lot of work with images.

So Yan is not only a pioneer but a leader in A.I.

research and implementation OK.

In the next lesson we will outline the solution to the M.S.

problem.

Thanks for watching.

مشارکت کنندگان در این صفحه

تا کنون فردی در بازسازی این صفحه مشارکت نداشته است.

🖊 شما نیز می‌توانید برای مشارکت در ترجمه‌ی این صفحه یا اصلاح متن انگلیسی، به این لینک مراجعه بفرمایید.