Machine learning terms you must know about as a beginner.

🧵👇

There are a million machine learning tutorials and courses out there which makes it really hard to choose the right one for you.

Here are some that I really like that you might find useful.
🧵 👇🏻

Before we begin, here's some unsolicited advice.

It is not about how many courses you take that matters but the problems that you solve with your knowledge.

You can take any combination of courses and tutorials from this list that you think will be useful.

Python crash course by Traversy Media (YouTube)

• Introduces you to a lot of basic concepts in Python
• Only 1.5 hours long

Intermediate Python course (YouTube)

• Helps you understand some of the more difficult concepts in Python

The neural networks series by 3blue1brown (YouTube)

• Visually explains neural networks
• Short and concise

Everyone tells you how important math is in machine learning.

But I believe that if you want to learn machine learning today, there are things more important than math which very few people will tell you about.

Here are 5 of them.

🧵 👇🏻

The field of machine learning has been very academic which is why there is so much emphasis on learning math for it.

But, today with the frameworks that we have, not knowing how to multiply matrices is far less of an issue than knowing how to use the tools in this list 👇🏻

Yes, I know; you have R, Julia, and whatnot for machine learning but as of today, Python is the industry standard.

Get your data structures and algorithms right in Python because you will need them, here are some concepts worth learning 👇🏻

Let's say you're building the next big GPT-3 powered app that's going to take over the world.

You add a new feature to your app but it breaks the entire thing, what do you do now?

Learn Git, the ultimate version control tool.

I use this cheat sheet for all my git needs 👇🏻

You will be constantly be working on multiple packages which specific python version requirements 🐍

Virtual environments like Anaconda or Poetry will make this process much easier for you.

Who likes to see broken dependencies while running code?

Machine Learning for the Web developer in 2021.

The beginner's guide.​

🧵👇

I started machine learning as a web developer, if I can do it then anyone can.

This carefully curated thread will give you key insights into my journey and how you can make this transition, seamlessly.

(2 / 14)

"Machine learning is not what you think it is"

One of the main reasons why people find it difficult to get started with machine learning is because of the lack of information, and rightfully so.

(3 / 14)

Machine learning as a concept has existed since the 1950s, but has only become popular in recent years because of the exponential rise of advancements in computer hardware.

(4 / 14)

In short, it because of the sudden rise of this technology ,which was previously unknown to the general public, that there is a lot of misinformation around it.

(5 / 14)

A git cheat sheet for all your needs!

https://t.co/H7QdOrthwV

This is how I got three job offers as a 16-year-old high school student.

( and why I rejected them )
🧵👇

June 2020.

I receive a DM from a college student in California. The message read something like this "Hey Pratham, I am one of the co-founders of XYZ Inc, a robotics company, we are looking to add enthusiastic people to our team, would you be interested?"

(2 / 11)

It was the first time I had ever been offered a job, so I said Yes! (I did reject it later)

There were a couple of interview rounds I had to go through. None of them included technical questions, which was interesting.

(3 / 11)

After the first interview, I saw the list of the roles I could apply for. I realized how little I knew, terms like "Ui/UX, ROS(Robot operating system)", at that time, I was overwhelmed.

(4 / 11)

I knew that this wasn't a job for me. I decided that I would reject this job role even if I cleared the second round, which I ended up doing.

(5 / 11)

Important paper from Google on large batch optimization. They do impressively careful experiments measuring # iterations needed to achieve target validation error at various batch sizes. The main "surprise" is the lack of surprises. [thread]

https://t.co/7QIx5CFdfJ

The paper is a good example of lots of elements of good experimental design. They validate their metric by showing lots of variants give consistent results. They tune hyperparamters separately for each condition, check that optimum isn't at the endpoints, and measure sensitivity.

They have separate experiments where the hold fixed # iterations and # epochs, which (as they explain) measure very different things. They avoid confounds, such as batch norm's artificial dependence between batch size and regularization strength.

When the experiments are done carefully enough, the results are remarkably consistent between different datasets and architectures. Qualitatively, MNIST behaves just like ImageNet.

Importantly, they don't find any evidence for a "sharp/flat optima" effect whereby better optimization leads to worse final results. They have a good discussion of experimental artifacts/confounds in past papers where such effects were reported.

1/ @werat asked about whether the debugger was using the C# compiler or language service in VS 2002. It was not. The debugger has a component called an ‘expression evaluator’ that is provided per language and is responsible for parsing and evaluating expressions when stopped at a

2/ breakpoint. For example, if you type into the immediate window, hover over a variable, type into the watch window, etc. the expression evaluator is involved. The debugger and the language service are actually deeply integrated in a number of scenarios in VS, which may

3/ initially seem surprising. I may talk about more of these scenarios in the future, but to give a flavor, when you set a breakpoint at design time the language service is involved, when you are using Edit and Continue the LS is involved, the range of what is being evaluated

4/ when you get a debugger data tip, completion lists when typing in the data windows or immediate window, etc. Prior to VS 2002, there were distinct debug environments tailored to each target customer (VB/VBA, C++, VJ, VBScript, TSQL, FoxPro, Fortran, etc.). The goals of the

5/ VS 2002 debugger included unifying the debugger backend components and providing a single UI stack that would enable new scenarios (for example, a callstack across languages). Expression evaluators are the architecture that separate the language specific pieces from the