Machine Learning Newsletter April 2021 1.5
Our best, content filled newsletter yet. Enjoy
Freebies and Easter Eggs for Pros
Found free ebooks for pro subscribers. We frequently share links to high quality free books found on the internet (legit ebooks from official sources). This month it is a database ebook. Learn about the future of database and algorithms. (pro link) Hopefully these easter eggs and freebies make your paid subscription worth extra.
Machine Learning Course: Important Formula of Machine Learning
Manhattan and Euclidean Distance explained. Pro member can get free access to this article. Message us hi@uniqtech.co Understand Manhattan Euclidean Distance in Minutes (medium)
Where does this important distance formula fit in the machine learning workflow? The answer is the evaluation phase of the workflow.
There are too many things to learn in life. Don’t know where to start? Uniqtech’s tutorial starts with a relevant current topic, our skill cards fill in the knowledge gap, helps you out as you read through trending articles. Think flash cards for machine learning. Sign up on https://ml.learn-to-code.co Substack paid subscribers here get access to pro level free.
Crypto Knowledge
We do not know much about crypto but here are some fun facts for informational purpose only. Crypto currency is highly risk. We are not financial advisors and we don’t endorse any coins. Our staff may own crypto coins.
Did you know that Dogecoin is a fork of Litecoin? The work is related, but Dogecoin also differs from litecoin. Altcoin short for alternative coins refer to coins that came after Bitcoin, and or Ethereum. A blockchain is a large public ledger. A miner’s job is to handle proof-of-work, and get rewarded for decipher the work: encrypted messages and transactions and add the next block to the blockchain. Mining requires specialized equipment such as ASIC devices. The crypto craze might have caused a shortage in NVIDIA graphic cards at one point, though it is not a specialized device.
Our upcoming python interview guide will explain blockchain and cryptography on a basic level, as interview questions. (Next next newsletter coming within 1.5 weeks.)
Perhaps an effort to capture the crypto mining market and allows gamers to purchase graphic cards without competition from miners, NVIDIA finally released a specialized card. NVIDIA CMP HX Dedicated GPU for professional mining https://www.nvidia.com/en-us/cmp/
Talking about NVIDIA, it also offers neural sticks : small devices that can perform complex deep learning tasks. The Jetson Nano Developer Kit is quite affordable.
Learn more here. (open access link) Learn about its performance here.
We will discuss the fundamental, technical knowledge of cryptocurrency and neural sticks in details in the near future.
Want to see the a toy example of blockchain explained? Message us. We will provide you with a flash card. https://ml.learn-to-code.co/message3.html
Deep Learning Workflow
To study machine learning or deep learning it’s first important to understand this workflow (map). It’s important when learning, to constantly check which step you are at. Always keep this workflow chart handy.
Where does this important distance formula fit in the machine learning workflow? The answer is the evaluation phase of the workflow.
Above you saw our illustration of the workflow. If you look at google cloud’s machine learning workflow illustration you will notice generally the ML providers agree with the same workflow.
Our machine learning course will also build on top of this flow.
End to End Machine Learning Course
In today’s lesson, we will explain a few common, important libraries, algorithms and models used in Machine Learning and Deep Learning. You have seen them, heard of them, now let’s understand them on a slightly deeper level, add some labels and categorize them.
General purpose tools: PIL for image processing, short for Pillow. CV opencv also an image processing library. skimage etc.
Image segmentation models: U-Net, Mask R-CNN is an improved R-CNN (Region Based Convolutional Neural Networks (R-CNN)).
Object detection methods: YOLO, RetinaNet.
The image below illustrate both computer vision tasks: segmentation and object detection. See that each person is blotted with colors - segmented from the background of the image. Each person is also identified - object detection, and marked with a bounding box. You can guess the computer vision task by looking at the images in the technical paper. How cool is that.
Finally the infamous generative adversarial networks (GANs) - the architecture behind deep fake videos. But let’s not focus on that. It is perhaps the coolest invention ever. GANs can now generate high definition, photo realistic people. Check it out https://thispersondoesnotexist.com/ By the way, living in Silicon Valley, we got to take selfies with Ian Goodfellow, the inventor. It’s cool to run into a genius. He’s fairly distinctive, and he was wearing a t-shirt : YES WE GAN (parody after the Obama presidential campaign slogan).
Virtual Company Office Tour
Too bad the pandemic has completely interrupted our work life. It still pays to be a developer, especially in Silicon Valley. We still want to highlight some developer lifestyle, digital nomad lifestyle, co-working spaces and tech office for you.
We all know that tech companies offer many perks and pamper their employees. Just how crazy can it get?
Need a new keyboard or new mouse for your next coding project? Facebook has vending machines that vends apple hardware and ergonomic mice. Silicon Valley startups even have Yokai Ramen machines right in their lobby. All in the name of keeping you working there longer.
Did you know that Airbnb has booths that are miniature of their popular bestselling airbnb rooms? Github’s office is filled with Octocat inspired pictures (see image below). We all know that Facebook murals and arts can be worth more than a million.
Facebook has la marzocco coffee machines serving Pytorch logoed coffee with latte art. Airbnb has 100 options of carbonated drinks, teas and beer on tap (See picture below). Accenture SF office has the European mega Nespresso machine - super fancy automated commercial Nespresso.
Like what you see? Let us know so we can post more of it.
In the next newsletter we will cover being a digital nomad traveling to Italy to attend Python conference and teach machine learning.
Trends and News
Toyota acquires Lyft’s auto driving unit. We summarize the key points and take-home message in this flash card (all access).
#MachineLearning trend #GPT3 partners with #Microsoft
https://ml.learn-to-code.co/skillView.html?skill=dp9SKpA4Lp6T4DsFkOJW
Learn more about the details here.
What if you have no access to #GPT3 #NLP what would you do? (pro member only) https://ml.learn-to-code.co/skillView.html?skill=jaKmxoiP9fWUzuDsMV7N
How is #DeepLearning is used in #Tesla #AutonomousDriving https://ml.learn-to-code.co/skillView.html?skill=Uyu1HHTe627tF2ffXCX0 #selfdriving #selfdrivingcars
Is tech exodus happening? Are tech companies leaving Silicon Valley and heading to Seattle, New York and Austin? How many left? Wait, Microsoft just bought a campus in San Jose. Check out this tech group's Tech Exodus tracker dashboard.
Data Science Development Environment Anaconda, Miniconda
Anaconda versus miniconda explained (pro access link)
Anaconda pro tips. (all access link)
End to End Machine Learning Course - GPT-3
Our developer got GPT-3 access. What is a question you want to ask GPT-3? Because GPT-3 is in a testing phase, some of the generated answers are not recommended to be shared. But what’s really cool is to understand how to design apps for such a new tech paradigm. Let’s discuss that today.
Our core developer’s personal GPT-3 testing / research goal is test multilingual natural language processing (NLP) using GPT-3 specifically East Asian languages, technical writing in Chinese, Japanese etc.
It’s really cool to think of GPT-3 as a brand new API which deserves brand new app design by developers. We will discuss more our pro access flash cards : how to design apps for GPT-3 (super PRO access, premium transcript). Any pro members can view our detailed transcript and notes on GPT-3 app design here. GPT-3 basics are available for all access. Just log in using your gmail to read on https://ml.learn-to-code.co
PRO members you may need to clear cache, re log-in to view the latest content on the landing page. A lot has changed.
Our GPT-3 landing page. https://ml.learn-to-code.co/path3.html?tutorial=gpt3_partial
GPT-3 alternative for “the poor man” (people without access) (pro member link). Here’s another alternative (pro member link)
GPT-3 Microsoft partnership. OpenAI licenses GPT-3 technology exclusively to Microsoft (all access link).
Did you know that Elon Musk runs some of the world’s most valuable companies such as : Tesla, SpaceX, and OpenAI. Get some insight on how is deep learning used at Tesla here (all access link).
In a future ML lesson, we will discuss transformer networks, including GPT-3. In the next next newsletter let’s discuss some cool use and application of GPT-3.
Next newsletter: coming out in the next few days easy NLP with Facebook.
What would a data scientist do with Dogecoin hype? Did you know you can easily analyze the Dogecoin blockchain? Or at least take a look at it.
Technical interview tips. Deep Learning formula explained series part 1.
Comparison of tech job positions.
Next next newsletter: Google I/O is happening soon, virtually. We will be publishing transcripts and notes of Google Cloud and Machine Learning at Google. There will be both basic access and pro access content available. Stay tuned.