Self-Supervised Learning [Includes SimCLR + Python Coding]

本课程教您“自我监督学习”（SSL），也称为“表征学习”。

SSL 是机器学习中一个相对较新且热门的主题，用于处理带有有限标记数据的存储库。

有两种通用的 SSL 技术，对比和生成。本课程的重点仅是有监督和无监督的对比模型。

本课程中有几个示例和实验，可让您完全掌握 SSL 背后的理念。

我们关注的领域是图像领域，但您可以将学到的知识应用到其他领域，包括时间记录和自然语言处理 (NLP)。

在每堂课中，您都可以访问相应的 Python .ipynb 笔记本。笔记本电脑最好使用 GPU 加速器运行。观看以下讲座了解更多详情。

如果视频太快或太慢，您可以随时更改它们的速度。您还可以打开视频字幕。

最好以 1080p 画质观看本课程的视频并打开字幕。

这些讲座的创建是为了在带有 GPU 加速器的 Google Colab 上发挥最佳效果。

这些讲座中使用的 TensorFlow 版本是“2.8.2”。您可以在 Python 笔记本的第一个单元格中使用 %tensorflow_version 2.x。

Python 中的机器学习库（包括 TensorFlow）正在不断发展。因此，您必须随时了解更改并修改您的代码。

课程大纲

四节十讲

第 01 节：简介。

第 1 课：课程介绍。

第 2 讲：Python 笔记本。

第 02 节：监督模型。

第 3 讲：监督学习。

第 4 讲：迁移学习和微调。

第 03 节：标记任务。

第 5 讲：标签挑战。

第 04 节：自我监督学习。

第6讲：自我监督学习。

第 7 课：监督对比借口，实验 1。

第 8 讲：监督对比借口，实验 2。

第 9 课：SimCLR，一种无监督对比借口模型。

第 10 讲：SimCLR 实验。

Self-Supervised Learning Including Supervised & Unsupervised (SimCLR) Contrastive Learning

Published 06/2022
MP4 | Video: h264, 1280×720 | Audio: AAC, 44.1 KHz, 2 Ch
Genre: eLearning | Language: English + srt | Duration: 10 lectures (4h 23m) | Size: 2.1 GB

What you’ll learn
Self-Supervised Learning Techniques
Supervised & Unsupervised Contrastive Learning
SimCLR Model of Chen et al. (2020)
Examples of Self-Supervised Learning in Image and Temporal Data Using TensorFlow in Python

Requirements
Python 3+ Programming Skills
Machine Learning Knowledge
Familiarity with TensorFlow 2.X
A Gmail & a Web Browser (Preferably Google Chrome)

Description
“If intelligence is a cake, the bulk is self-supervised learning, the icing on the cake is supervised learning, and the cherry on the cake is reinforcement learning.”

Yann André LeCun

Chief AI Scientist at Meta

Some “Musts” Before Starting

You must be familiar with deep learning architectures, including stacks of convolutional, recurrent, dense, pooling, average, and normalization layers using the TensorFlow library in Python 3+.

You must know how to develop, train, and test multi-layer deep learning models using the TensorFlow library in Python 3+.

You must know that this is a “100% Money Back Guarantee” course under Udemy rules.

Course Instructor

My name is Mohammad H. Rafiei, Ph.D. I am honored and humbled to serve as your instructor.

I am a machine learning engineer, researcher, and instructor at Johns Hopkins University, College of Engineering, and Georgia State University, Department of Computer Science. I am also the founder of MHR Group LLC in Georgia.

Subject & Materials

This course teaches you “Self-Supervised Learning” (SSL), also known as “Representation Learning.”

SSL is a relatively new and hot subject in machine learning to deal with repositories with limited labeled data.

There are two general SSL techniques, contrastive and generative. This course’s focus is on supervised and unsupervised contrastive models only.

There are several examples and experiments across this course for you to fully grasp the idea behind SSL.

Our domain of focus is the image domain, but you can apply what you learn to other domains, including temporal records and natural language processing (NLP).

In every lecture, you can access the corresponding Python .ipynb notebooks. The notebooks are best to be run with a GPU accelerator. Watch the following lecture for more details.

If the videos are too fast or too slow, you can always change their speed. You can also turn on the video caption.

It is best to watch the videos of this course using 1080p quality with the caption on.

The lectures are created to work best on Google Colab with GPU accelerators.

The TensorFlow version used in these lectures is ‘2.8.2.’ You may use %tensorflow_version 2.x at the very first cell of your Python notebook.

Machine learning libraries in Python, including TensorFlow, are evolving. As such, you must keep yourself updated with changes and modify your codes.

Course Overview

Four Sections and ten Lectures

Section 01: Introduction.

Lecture 01: An Introduction to the Course.

Lecture 02: Python Notebooks.

Section 02: Supervised Models.

Lecture 03: Supervised Learning.

Lecture 04: Transfer Learning & Fine-Tuning.

Section 03: Labeling Task.

Lecture 05: Labeling Challenges.

Section 04: Self-Supervised Learning.

Lecture 06: Self-Supervised Learning.

Lecture 07: Supervised Contrastive Pretext, Experiment 1.

Lecture 08: Supervised Contrastive Pretext, Experiment 2.

Lecture 09: SimCLR, An Unsupervised Contrastive Pretext Model.

Lecture 10: SimCLR Experiment.

Who this course is for
Machine Learning Students and Enthusiasts
Those Want to Learn Self-Supervised Learning and Practice It in Python