Basic Intuitions of Machine Learning & Deep Learning for Beginners

Before we start, here is a public announcement from the author: The earlier editions of this article had some link issues. If you encountered some unpleasant experiences, I would apologize for that. Now fixed and clean. Please enjoy exploring the agenda.

Deep Learning is a data hungry, energy hungry, and often precarious monster ('Underspecification' problem). In the past, success stories of Machine Learning—especially those of Deep Learning—have been over-emphasized. In recent years, increasing number of its shortcomings have been acknowledged in the public domain.

In order to understand both its weaknesses and its strength, we need to have basic understanding on how Machine Learning and Deep Learning operate.

This series is intended to give beginners a preview of the topic inside out to demystify the hype and myths behind the subject.

You can continue the content of this series from the links here. All free of charge:

• Introduction, Artificial Intelligence, Machine Learning, & Deep Learning: This chapter explores distinctions among three terminologies—AI, Machine Learning, and Deep Learning—and shows a quick overview of their historical background
• Chapter 1: Machine Learning Algorithm Paradigm: To a great extent, the limitation of the traditional algorithm paradigm set the stage for the emergence of Machine Learning. Machine Learning emerged in response to the limitation of the traditional algorithm paradigm. What is Machine Learning Algorithm Paradigm? How is it different from the Traditional Algorithm Paradigm?
• Chapter 2: Generalization, the Ultimate Goal of Machine Learning Project: The ultimate goal of Machine Learning is Generalization. Well, is Generalization achievable? The question causes perennial enigma to users. After all, Deep Learning is as good as the given dataset. This chapter also illustrates how a Deep Learning pipeline is structured in an attempt to generalize its output.
• Chapter 3: Deep Learning Connectionism Foundation—Philosophy of Connectionism: Deep Learning is inspired by the philosophy of Connectionism. And Connectionism speculates: while a single neuron is very simple, when multiple neurons are connected as a neural network, they can collectively exhibit very intelligent behaviours. This chapter illustrates how Connectionism influenced the architecture of Deep Learning.
• Chapter 4: Deep Learning’s Learning Mechanism—Optimization Paradigm driven by 3 Step Iteration Cycle: 1.    How does Deep Learning learn? Simply put, Deep Learning is an Optimization Paradigm. It operates repetitions of 3 Step Operation Cycle: Try, and Error, and Refine. This chapter will show you the learning mechanism of Deep Learning: how it operates its optimization iteration cycle.
• Chapter 5: Deep Learning Revolution—Engineering Achievement on Steroid: In the past, the advancement of Deep Learning was hindered by hardware constraints. Today, the explosion of hardware computation power has accelerated the advancement of Deep Learning. In a nutshell, Deep Learning is an engineering achievement on the steroid of computational power.
• Chapter 6: Deep Learning’s Carbon Footprint—Imperative for Energy Efficiency Revolution: Deep Learning can be an Energy Hungry monster. Today, some advanced Deep Learning models demonstrate remarkable achievements that no conventional algorithm method has failed to achieve. Nevertheless, as its flipside, some successful Deep Learning applications consume an enormous amount of energy. Carbon Footprint of Deep Learning needs to be addressed in the age of Climate Change, our own existential crisis.
• Chapter 7: Deep Learning’s Blind Spots —Underspecification: “Underspecification is ubiquitous in modern applications of ML, and has substantial practical implications”, according to a paper released by a group of Google Data Scientists. An ML pipeline is underspecified when it can return many predictors with equally strong performance in the training domain (in the lab). When those equally validated predictors in the lab are deployed in real world settings, they can demonstrate different sensitivities to changes in conditions outside the lab. Underspecification could lead to ‘model instability’. When deployed in some real world applications, a model generated from an ‘underspecified’ ML pipeline can result in detrimental consequence involving human lives.
• Chapter 8: Deep Learning’s 2 Basic Models: Sequence Models and Convolutional Neural Networks: Although Deep Learning belongs to the family of Machine Learning, an algorithm paradigm that leans without being explicitly programmed, it demands a certain level of specifications. And Data Representation and Model Specifications are very important to design a successful Deep Learning application. Here, we can have a quick preview of some features in Data Representation and Model Specifications of 2 basic Deep Learning Models: namely CNN and Sequence Model.

Get Right Intuitions

In retrospect, at the beginning it was not easy for me to grasp the high-level conceptual framework of Machine Learning and Deep Learning. Whatever the real reason it might have been, to get right intuitions and inspirations was very helpful for me to digest the complexity of the subject.

Now, reflecting my own experience, my intention for this series is to share those intuitions with beginners. The primary purpose is to give them a good starter to digest the upcoming full course menu of these complex subjects. In this context, as a general policy throughout this series, I would like to draw a rough guiding sketch rather than to go into details. Fortunately, there are oceans of open-resources that you can dive deeper into for these topics. Instead of creating redundant works, my intention in this series is to present guiding intuitions that help the readers navigate their own Machine Learning & Deep Learning journeys.

I hope that you find my project helpful for your Machine Learning journey.

More than anything, enjoy your journey!

Thank you.

Best Regards,

Michio Suginoo

https://basno.com/qsouz72l

https://www.youracclaim.com/badges/e387faef-cb8b-4600-aebc-8a83645acb90

https://www.coursera.org/account/accomplishments/specialization/B4NLZ3ATYT87