Beginners guide to start with machine learning.

What you need:

1. A programming language
2. A place to write and run code
3. A way to deal with data
4. A way to visualize results
5. A cool algorithm
6. An introductory course
7. A project to solve

Here are my recommendations.

Most people recommend Python to start.

I do too.

You can do machine learning with many different languages, but today, Python is the best option.

My advice: Get comfortable writing code before looking into machine learning.

You don't have to be a great developer to start.

But if you are, it helps.

Everything you know about building software is helpful to build machine learning systems.

Good developers have the odds in their favor.

"How much Python do I need to know before starting?"

If you ask, you aren't ready yet.

Ideal scenario: You should be comfortable building software. Most people get here within a year.

Let's move on to the 2nd item from the list: A place to write and run your code.

Get familiar with Google Colab.

• It's free.
• Requires no setup to start.
• It's available from anywhere.

Alternative: Kaggle.

20 machine learning questions that will make you think.

(Cool questions. Not the regular, introductory stuff that you find everywhere.)

🧵👇

1. Why is it important to introduce non-linearities in a neural network?

2. What are the differences between a multi-class classification problem and a multi-label classification problem?

3. Why does the use of Dropout work as a regularizer?

👇

4. Why you shouldn't use a softmax output activation function in a multi-label classification problem when using a one-hot-encoded target?

5. Does the use of Dropout in your model slow down or speed up the training process? Why?

👇

6. In a Linear or Logistic Regression problem, do all Gradient Descent algorithms lead to the same model, provided you let them run long enough?

7. Explain the difference between Batch Gradient Descent, Stochastic Gradient Descent, and Mini-batch Gradient Descent.

👇

8. What are the advantages of Convolution Neural Networks (CNN) over a fully connected network for image classification?

9. What are the advantages of Recurrent Neural Networks (RNN) over a fully connected network when working with text data?

👇

What's machine learning?

Not a framework. Not a different language. Not the latest "cool" thing around the Internet.

Thread: Let me convince you that machine learning is the best thing since sliced bread 🍞.

Let's start with a simple exercise:

"Write a function that returns 0 if a supplied number is even or 1 if the number is odd."

If you are a developer, I'm sure you know how to write this function. It should be simple, right?

To avoid using a specific syntax, let's represent this function as follows:

y = f(X)

Where:
• X → The input number
• y → The result (0 if even, 1 otherwise)
• f → The function that does the work

We can now get to the interesting part.

If you are a programmer, writing this function is simple.

You know the relationship between X (the input number) and y (the output), so you can easily write the rules (the function.)

In machine learning, we don't write the rules. We learn them.

Imagine we had an algorithm that processes many samples (X → y), learns the relationship between them, and comes up with the correct rules.

We call these samples "training data," and they would like like this:

• 2 → 0
• 5 → 1
• 8 → 0
• 4 → 0
• 3 → 1
• ...

I'll blow your mind with a technique you aren't using yet.

Sometimes, you want your system to do exactly the opposite of what your machine learning model thinks you should do.

Let me convince you. ↓

I'm going to start with a nice problem:

Imagine a model that looks at a picture of an electrical transformer and predicts whether it's about to break or not.

Don't worry about how the model does this. We are going to focus on the results instead.

There are 4 possible results for this model:

1. It predicts a bad unit as bad.
2. It predicts a bad unit as good.
3. It predicts a good unit as bad.
4. It predicts a good unit as good.

#2 and #3 are the mistakes the model makes.

Assuming we run 100 units through the model, we can organize the results in a matrix:

• The rows represent the "actual" condition of the transformer.

• The columns represent the "prediction" of the model.

We call this a "Confusion Matrix."

This is how we can read this confusion matrix:

• 60 bad units were predicted as bad.
• 3 bad units were predicted as good.
• 7 good units were predicted as bad.
• 30 good units were predicted as good.

A surprising situation.

A machine learning model predicting 3 classes ends up with these results:

• Class A: 90% accuracy
• Class B: 80% accuracy
• Class C: 70% accuracy

Class C seems to be the worst-performing.

But you know already that I'll say it isn't. ↓

Here is the correct answer: We don't know yet which class is the worst.

This is important to understand, so keep reading.

Many people make this same mistake every day and try to fix things that ain't broken.

We are missing something, so let's go and get it.

How about if we ask a friend to look at the dataset manually?

Let's say these are the results of our friend doing the job:

• Class A: 95% accuracy
• Class B: 100% accuracy
• Class C: 71% accuracy

Do you see what just happened?

The model is actually really accurate when predicting Class C! 70% versus 71% when done by a human.

Class B, however, is a disaster: 80% versus 100% when done manually.

Class B seems to be the worse performing class!

The work that our friend did is called a "baseline."

Without having it, it's hard to determine how good are the results of a model.

I've seen tons of money down a drain just because teams didn't create a baseline beforehand.

But now, that won't happen to you 😉.

The 8-step quick-start guide to learn Machine Learning.

🧵👇

1⃣ Start with Python 🐍

Yes, you can do other languages, but Python is by far the most straightforward option.

👇

2⃣ Get familiar with numpy, pandas, and matplotlib

These three libraries are probably the most common Python libraries you'll have to use every day.

(Even if you don't end up doing machine learning, these libraries are awesome and useful.)

👇

3⃣ Start using notebooks

Look into Jupyter or Google Colab.

Notebooks are essential for data scientists and machine learning practitioners. Most of the code you'll read and write will be in notebooks.

👇

4⃣ Find a problem (already solved)

In my opinion, the best way to start is by working through a problem —especially when you can learn from its solution.

Start with something simple. I usually recommend "Titanic" from Kaggle.

👇

I learned enough AI/ML in 12 months to change my life forever.

You can do the same, and it all starts with a killer process to learn new things.

You can steal my approach. Every detail on the latest issue of my newsletter: https://t.co/9gaenLByFb.

Make it work for you.

My recommendations to start with machine learning math:

• Statistics 110: Probability — Harvard University
• Essence of Linear Algebra — 3Blue1Brown
• Multivariate Calculus — Coursera

Links and 3 additional free resources here: https://t.co/fEmpmRKEvJ

If you are looking for more machine learning content, give me a follow and I’ll keep your feed interesting 😉.

My recommendation to start learning how to deal with structured data: Kaggle's Pandas Tutorial.

• Creating, Reading, Writing
• Indexing, Selecting, Assigning
• Summary Functions, Maps
• Grouping, Sorting
• Data Types, Missing Values
• Renaming, Combining

100% Python 🐍.

Machine learning education is broken.

If you are preparing for a research position, you are good. If you are looking to get out there and start solving problems, not even close.

Here are some thoughts so you can get ahead.

↓

Most classes, courses, and books cover the same road.

They start with a dataset. They finish with a working model. The focus is always on everything that happens in between.

Dataset → Model.

This is great, but not enough.

Real-life situations rarely start with a dataset, and they never end after you finish building your model.

Applying machine learning successfully is hard.

Here are a few examples that you should keep in mind.

First challenge: Properly framing up the problem.

If you don't understand the problem, you can't determine what data you need. If you don't understand the data, you can't build a good model.

I've never seen a company that had their data ready to go.

In fact, most of them don't even have data at all and need you to determine what exactly they should start collecting.

You usually have to go Problem → Potential Solution → Data.

11 key concepts of Machine Learning.

— Supervised Learning Edition —

🧵👇

😜

Before starting, remember that, if you follow me, one of your enemies will be immediately destroyed (and you'll get to read more of these threads, of course.)

And if you don't follow me, well, you just hurt my feelings.

😜

1. Labels

(Also referred to as "y")

The label is the piece of information that we are predicting.

For example:

- the animal that's shown in a picture
- the price of a house
- whether a message is spam or not

👇

2. Features

(Also referred to as "x")

These are the input variables to our problem. We use these features to predict the "label."

For example:

- pixels of a picture
- number of bedrooms of a house
- square footage of a house

👇

3. Samples

(This is also known as "examples.")

A sample is a particular instance of data (features or "x.") It could be "labeled" or "unlabeled."

👇

For a long time, I didn't understand how to use Virtual Environments in Python 🐍.

If this is just, let's end it here and now: 🧵👇

[2] Virtual Environments let you deal with the dependencies that your code has with external Python libraries.

It avoids having conflicts when your projects depend on different versions of the same library.

👇

[3] Let's imagine that you are building your first Python project and you install the "requests" library:

pip install requests

You get version 2.24.0 installed in your system.

👇

[4] A month later, you decide to work on your second project. It also needs the "requests" library.

But the latest version is not 2.24.0 anymore.

Now version 3 is available, and that's the one you want to use!

👇

[5] You could upgrade your entire system to version 3, but then you'll be potentially breaking the first project you built that depends on 2.24.0!

Can you imagine this happening on a server with many more applications running?

👇

An introduction to one of the the most basic structures used in machine learning: a tensor.

🧵👇

Tensors are the data structure used by machine learning systems, and getting to know them is an essential skill you should build early on.

A tensor is a container for numerical data. It is the way we store the information that we'll use within our system.

(2 / 16)

Three primary attributes define a tensor:

▫️ Its rank
▫️ Its shape
▫️ Its data type

(3 / 16)

The rank of a tensor refers to the tensor's number of axes.

Examples:

▫️ The rank of a matrix is 2 because it has two axes.
▫️ The rank of a vector is 1 because it has a single axis.

(4 / 16)

The shape of a tensor describes the number of dimensions along each axis.

Example:

▫️ A square matrix may have (3, 3) dimensions.
▫️ A tensor of rank 3 may have (2, 5, 7) dimensions.

(5 / 16)

I told everyone that I didn't care.

"Screw math! I've never been great with it, so I'm not starting with machine learning to fail at the end."

That was many years ago.

Math is still hard, but I don't think you should be scared at all. Here is why: 🧵👇

[1] One thing changed my mind: school pushed me to the deep end of the pool, and I was forced to swim.

I had to face my fears, and I started machine learning and realized that the math involved is not as scary (or as much) as some people believe.

👇

[2] Probably one of the most frequent questions I get is around the math needed for machine learning.

Answer:

▫️ Probabilities and Statistics
▫️ Linear Algebra
▫️ Calculus

But it turns out that this is not helpful.

👇

[3] What are you gonna do with this?

Are you going to learn all there's is about linear algebra before looking into machine learning?

How about probabilities and statistics? Or calculus? Do you know how extensive calculus is?

👇

[4] Many people stop looking into machine learning as soon as somebody tells them that calculus is a pre-requisite.

First of all, the amount of calculus you truly need is minuscule.

Second, without proper direction, there's no way you'll find your way through the noise.

👇

10 machine learning questions you should be able to answer.

I started experimenting with a questionnaire before interviewing candidates. Applicants submit a video or written answers to apply for the position.

Thread: Let's look at 10 of these questions.

The questionnaire solves a couple of problems:

1. Weeds out candidates that aren't ready for the job.

2. It significantly cuts down the time we spend talking synchronously.

One problem: It requires extra time from the candidate before the phone interview.

Here are some numbers on a job posting before using the questionnaire:

• 249 applicants
• 13 good applications

That's only 5%.

This sucks.

Numbers after using the questionnaire:

• 44 applicants
• 21 good applications

That's 47%.

This is much better.

I look at it this way:

Sooner or later, I'll ask these questions, and the candidate will have to answer them.

I'd rather give them space and the opportunity to think about their answers without any pressure.

Those who can't or don't want to answer won't apply. That's fine.

A topic that comes up in every interview.

Bias, variance, and their relationship with machine learning algorithms. One of the most basic concepts that you have to know by heart.

Here is a simple summary that you will easily remember.

↓

Every machine learning algorithm deals with 3 types of errors:

1. Bias error
2. Variance error
3. Irreducible error

There's nothing we can do about #3.

Let's focus on the other two.

↓ 1/5

"Bias" refers to the assumptions the model makes to simplify the process of finding answers.

The more assumptions it makes, the more biased the model is.

"Variance" refers to how much the answers given by the model will change if we use different training data.

If the answers stay the same regardless of the data, the model has low variance.

Often, linear models are high-bias, and nonlinear models are low-bias.

Example low-bias algorithms:
• Decision Trees
• SVN
• kNN

Example high-bias algorithms:
• Linear Regression
• Logistic Regression