This is an email from Machine Learning Bytes, a newsletter by Data Science Bootcamp.

Hello 2022 Year of AI and Blockchain Data, Year of Analytics!

A big UIUX re-design is coming soon to our site, the new course panel format will make the content much easier to navigate and consume. This is another long newsletter. Don’t be overwhelmed by the number of flash cards. They will be organized into easily consumable smart tutorials. We will have opportunities to see how the concepts are useful in various different contexts. They will make appearances in projects (coming soon 2022).

Today’s newsletter covers the training loop and training step of machine learning. It’s an important topic! We also cover blockchain web3 recruiting and basics. What is avalanche, polygon, alchemy?? And finally we sprinkle the newsletter with a few technical interview tips including Google’s guide to python, Steve Jobs explaining OOP in plain English. Have you made learn-to-code, learn machine learning your goal for 2022? There’s still time to make that your new year resolution!

  • [pro] = flashcards for pro, paid members only.
  • [pub] = flashcards temporarily available for all to view

More payments and tipping options are now available. Become a paid member today! Check out the pricing payment page.

Mini projects, lessons, technical interview lessons and developer NFT launching 2022. Wow how come we only charge $5/month?! Like all startups, we want to be so good that many of you will subscribe, so we end up charging very little per person. And plus, we are investing in software + tech > overhead, so we continue to be a lean startup. These projects will be beginner friendly. We aim to have projects that take only a few lines of code but can demonstrate big concepts. You don’t have to worry about another certificate that you can’t finish. We have been there, we understand!

Substitute this meme with any other course platforms: Udacity, Coursera, …. even our founder has unfinished business in them all!


Vote for any flash cards you like by clicking the vote button at the bottom of each skill card.

Your votes help our algorithms understand what to recommend.


The best value of our newsletters is the high quality freebies (easter eggs) we find on the internet, such as high expensive ebooks and courses available for free for a limited amount of time. [pro]

[Value Proposition]: If there are great courses, by top professors for free, then why do you still need us? We take copious notes of deep learning concepts, transcribe important talks, collect jobs data, make flash cards. We then compile them into meaningful, cohesive courses, and smart tutorials. We also cover current events, hiring trends, tech trends, webinars and talks. We also provide insights on the recruiting, technical interview process, and the tech stacks and layers of your dream companies. We also link to more freebies and events. Don’t see something you need? Message us on the message tab, and or email to request content.


In this newsletter, we want to start tackling two complex topics in deep learning : training, autograd. These two topics open doors to a myriad of related topics: auto differentiation, gradient descent, gradient tape, training loop, back propagation etc. All these concepts are related, yet there are plenty of differentials. No pun intended.

Training dataset is used for training.

Cross validation dataset is used for evaluating the model, fine tuning the model parameters. The third dataset, aka the test dataset is for final model selection. Testing the model’s performance on unseen dataset (mimicking real world data), testing its ability to generalize.

Introduction to Machine Learning Training

Train test split (a review)

Training in Pytorch requires us to write a more custom detailed training loop. In scikit-learn, in tensorflow, training can be calling a simple high level API .fit(). Though each model has different architecture under the hood, the high level API has been abstracted to be .fit(). Here’s a fully annotated version of pytorch training loop. We will make a youtube video about it as well.

Train test split prevents overfitting


Train test split basics

train test split explained 02

Annotated training loop, pytorch source code [pro, IMPORTANT, high quality]

What is model.eval() [pytorch, pro]

Summary of Prof. Yann LeCun’s insights on deep learning training, optimization from the introductory deep learning class.

What is back propagation: calculate gradient of the loss function with respect to (w.r.t) parameters (weight parameters) of the neural network.

How is back propagation related to neural networks, weights, loss function, gradient descent, stochastic gradient descent, chain rule. [pub]

Common types of optimizers [pub]

A helpful analogy for understanding gradient descent [pro]

Understand the intuition behind gradient descent, and why it is challenging

Training step explained in details, gradient descent update function

Prevent overfitting in Machine Learning Deep Learning Models [pub, high quality]


[Joke, humor] Have you seen XKCD’s great comic about what machine learning is? check it out [pub, not affiliated]

Basic building block of a neural network, a basic neuron [previous newsletter]

What does this model do? [pro, high quality, cheat sheet]

Tensor definition 04 [pro, analogy, linear algebra] In the previous newsletter, we covered 3 definitions. Don’t worry you will see these in smart tutorials again.

Definition of Epoch [pub]

Set data type dtype in numpy


Machine Learning conferences you should know [pub]

Recurrent Neural Network (RNN) insights [pro]

Self driving car hardware, chips

Mini deep learning quiz, computer vision

Onnx model format defined. What is Onnx and why should you care?

Using GPU with CUDA in Pytorch [pub]



Web3 explained using a tweet

For general blockchain, crypto, web3 info, book mark our landing page (glossary vocab page) [log in to view, free, pub]

Previously we talked about the Work at Startup Expo at Y Combinator. Presenters including OpenSea and Coinbase. Check our previous newsletter for details and notes.

2022 just started and NFT market OpenSea already made many records! Why is OpenSea so hot? What is the tech stacks at OpenSea? Coinbase co-presented at the event. What is coinbase’s engineering goals? What are the problems most interesting to coinbase?

OpenSea recent funding (news)

Some even speculate that OpenSea is aiming for an IPO after hiring Lyft IPO CFO.

Coinbase basics:

How big is coinbase right now? Visit the link below to see its ecosystem size and composition.


Coinbase career site links:

Coinbase Engineering Lead on why it is not too late to get started with crypto, blockchain, and what kind of engineering problem coinbase wants to solve. [pro]

Coinbase Engineering lead on exciting technical challenges for coinbase

OpenSea: OpenSea software engineering job and tech stack, what qualifications do you need to have to work at OpenSea, the biggest, hottest NFT market space [pro] Carefully inspect and understand opensea’s tech stack at the end of the software engineer qualification section.

For more info on OpenSea, check out previous newsletter. Paid pro members just message or email us.


What is metamask? Metamask is a crypto wallet, the gateway to the blockchain, metaverse, NFT world, and many decentralized apps (dapps). Its logo is a 3rd fox.

[Project, web dev, crypto] How to inspect Tesla payment page to know what kind of payment options and crypto payment options it offers? How to check that Tesla may offer dogecoin payment for its car. [pub, project, mini project, important]

🏅 We aim to have more mini projects like this. Like it? Vote for it. Let us know

Important features of a dapp, decentralized app [pro]

What is alchemy,

What’s Avalanche (blockchain)?

What is polygon (blockchain)? flash card [pro]

Polygon + [pub]

What is IPFS? Define IPFS [pub] It’s likely that the NFT you or your friend purchased is hosted on IPFS.

What is ENS .eth?

If you see people using names such as .eth, they are using ENS service. Example Yourname.eth. Read more trivia like this on our blockchain landing page. You bet one of our staff has a .eth name. It’s hyped right now. ENS is “Decentralised naming for wallets, websites, & more.” If you see people using names such as .eth, they are using ENS service. Example Yourname.eth.


Did you know Google offers a free online tutorial for python? [pub]

Think about interviewing with Google using Python? You need this tool. [pro]

Getting started with python [pub]

Platforms for displaying your data science portfolio [pro]


Have you heard of Steve Jobs explaining the programming concept OOP? [pub]

Heap — FAAANG company’s favorite interview questions

Tree traversal, recursion with illustration [new and improved, pro, illustrated]

In-order Tree traversal, pseudo code and detailed explanation [pro, high quality, pro tip]


Next Newsletter we will start to cover the next step of machine learning workflow, while still adding info about training (because training is huge)! We will want to get started with two big fields computer vision and Generative Adversarial Networks (GAN). Both have interesting real world projects and applications. On the blockchain web3 side, let’s discuss how to use crypto wallets as login, online identity and basic wallet. That is a very useful part of web3 that is easy for web2 developers to start. There will also be useful mini projects in deep learning and data science in the 2022 newsletters. We will learn to build small useful projects. Also in our newsletter pipeline : Tesla AI Day transcript and notes. We will need to spend more time diving into the math behind gradient descent, including a review of chain rule and a recap of Andrew Ng’s formula for gradient descent.

We are beefing up our technical interview sections like tree based interview questions. What makes our interview prep a bit different from question bank sites like Leetcode and Hacker Rank is that we give you information to solve types of interview questions in groups instead of solving them one by one. There are too many questions, too many variations, we need to learn the patterns to be effective. Example: previous flashcard on how to solve matrix interview questions.

What’s amazing is that this framework works well for Uber Engineering and Google Cloud ML. Check out their ML workflow charts here. Machine learning workflow variations at Google and Uber Engineering. [uber] [google]

You can also find the link on our workflow landing page. ​​

Developer NFT designs coming soon

“Hmm something brewing: we may have NFTs available for learners to grab in 2022. The more you learn the cooler NFTs you get. What kind of developer NFTs will you get ;-) Paid members and donors get priorities. “

Everyone learning to code, learning machine learning is like Harry when he first heard about Hogwarts. It will be a magical journey from here. We will take some inspiration from this series to our developer NFT.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store