Tweet

This Thread may be Removed Anytime!

Twitter may remove this content at anytime! Save it as PDF for later use!

More from @svpino

Santiago

@svpino

1 Dec

What is machine learning and why you should care about it?

Let me try to convince you: ↓

2. Let's start with a simple programming 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.

3. To simplify things, let's represent this function as follows:

y = f(X)

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

We can now get to the interesting part.

Read 16 tweets

Santiago

@svpino

29 Nov

One of the things I love about Python: Duck Typing + EAFP.

A short thread that will change the way you write code.

↓

2. The idea behind Duck Typing:

If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.

In other words: the functionality of an object is more important than its type. If the object quacks, then it's a duck.

3. Duck Typing is possible in dynamic languages.

Look at the attached example.

Notice how "Playground" doesn't care about the specific type of the supplied item. Instead, it assumes that it supports the bounce() function.

Read 11 tweets

Santiago

@svpino

22 Nov

For one second, let's forget the idea of a central database.

Imagine a product that doesn't store your data. Instead, you keep that information. You allow others to use it at your convenience and close it whenever you want to.

Let's talk about one example. ↓

@kaggle

2. Think about those people that have built excellent profiles using @kaggle.

They have participated in many competitions, shared their knowledge, and built impressive curriculums.

What would happen if Kaggle decides to ban them?

3. This is not science fiction.

Every single company out there can ban you if they decide to do so.

Maybe it is justified, maybe it isn't, but that's beyond the point.

What happens then?

Read 18 tweets

Santiago

@svpino

19 Nov

Deploying a machine learning model is not a trivial task.

Here are some of the questions you may have to answer every time: ↓

1. What's the input format expected by your service?

2. How can we validate the input is valid? What's the appropriate action if it isn't.

3. What transformations are needed to turn the service's input into the model's input?

4. What transformations are needed to turn the model's output into the service's output?

5. Do we need to allow for batch processing of data?

6. How much time do we have to return an answer?

Read 6 tweets

Santiago

@svpino

10 Nov

Over the last few months, I've introduced three main improvements to how I build machine learning models.

Keep in mind that my job is focused on Computer Vision, and I mostly use TensorFlow and Keras.

Here are the highlights: ↓

First, I replaced image generators with the tf.data API.

This change alone has had a major impact on training time. But it doesn't stop there:

• My code is much cleaner
• A data pipeline is easily reusable

tf.data is a must.

Together with loading data, I used image generators to perform data augmentation.

Now, I try to make data augmentation part of the model using Keras' preprocessing layers.

These augmentations now happen in the GPU. This is another nice boost!

Read 6 tweets

Santiago

@svpino

8 Nov

If I were to start building a career in machine learning today, here is where I'd focus:

1. Python from the get go.
2. Learn how to build software.

I'd take my time here and avoid rushing into the "machine learning" specific stuff.

Something interesting happens here: ↓

A lot of people start learning software development because they want to get into machine learning.

Then they realize that machine learning is not what they care about.

This is great: there are many ways to build a successful career in the software industry.

As soon as you're comfortable, here is what I'd tackle next:

3. Machine learning fundamentals
4. Hands-on machine learning

I like to cover these at the same time, instead of one after the other: learn some theory, then apply it right away.

Something to keep in mind:

Read 12 tweets

Share this page!

Santiago

Try unrolling a thread yourself!

More from @svpino

Santiago

Santiago

Santiago

Santiago

Santiago

Santiago

Did Thread Reader help you today?

Like this author's thread?