Solving Machine Learning Problems – Falling in reverse without realizing it.


Disclaimer: this post is a summary of Chapter 1: Falling in Reverse, to Swim or Sink, from my upcoming book: Machine Learning For Bearded Hipsters. You can support the project here.


A common thing when you are falling in reverse is to don’t realize it by yourself. All your friends and close people are foreseeing your catastrophe but you are so happy petting your pink unicorn that you don’t realize how bad the situation is until it slaps your face.

falling

You are not the only one on this but its really helpful both in Machine Learning and Life to know when you have a problem and the nature of it, so you can help yourself by asking:

  1. Do I have a problem?
  2. What is the Nature of my problem?
  3. What is a valid solution to my problem?

The problem of living in big cities

big-city

For example, when you are looking for a flat to rent, you need to know how difficult and expensive is it to find, how much time in average does it take to start the search until you finally move. So your problem will be to find a proper flat due to a determined date, if you don’t have a date, then is not a problem, is your curiosity trying to make the decision.

Next, you need to define what kind of flat you want, what kind of flat you need, and which sacrifices will you do at each side to make the intersection of them bigger. So you increase your possibilities of success. If you don’t have a valid solution or what you need as a minimum, you will fail, since what you will do is to find the cheapest or the fanciest flat without taking into account what really matters to you.

STEP 1. I have a problem.

problem-car

Like in drugs or alcohol, in modern living and machine learning you also need to admit that you have a huge ugly pimple in your … face.

STEP 2. What my problem is exactly?

exact-probrlem

Try to describe your problem in one sentence. Later, try to get the best of it adding a list of details your problem has or you should take into account. Also, define which data or medium are you using to solve the problem.

To solve this I like to use a resource I learn working in the media, the 6W method. What? Where? Why? When? How? Who?

What? The what represents the description of the problem itself. For example,  I need to find a room in Barcelona area.

Where? Represents the medium and data you will use, where you will find the knowledge. For example, I’ll find in Idealista and Fotocasa or other online house renting platforms.

Why? The why is so obvious that sometimes we forget to define it. But is necessary since it gives the motivation for the problem. For example, if I do not find the flat, I will sleep in a bank office folded with dirty carton.

When? Is the amount of time you have to find a solution. For example, you know that some methods like Deep Learning require more time than traditional methods so you need to consider the amount of time available vs the amount of time consumed by your method. In the case of finding a flat, it would be, for example, 30 days.

How? Is the definition of your action plan itself, how will things be done. Set a to-do list with all the tasks and start working on them. For example, I’ll train a multivariable regression method with my preferences and the characteristics of the flats so it sends me a notification with the proper ones for me.

Who? If you are a team, it makes sense to define who will work on each of the previous tasks. If you are working alone, you need to find which tasks you will be able to do and in which you may need extra help. Also, identify who will be able to help you with these tasks. Plus check how others are solving similar problems. For example, if I’m not able to find the flat in the first 10 days and I start feeling neurotic, I’ll hire an agency to help me.

The minimum viable solution

shake-hands

As well as define the problem, it is also important to define success. What would be a successful solution? To have a 90% accuracy? To improve state-of-the-art model X.

As a minimal solution, it is important to state what can you give up with and what you cannot sacrifice. Everything more than this minimum solution is our benefit, your own victory.

For example, I can pay at the most 400€ per month since I’m a predoctoral slave. I need to live in the city center but its ok if it less than 30 minutes far with public transportation. It is a must for the room to has a window but furniture is not necessary (I’m a very minimalistic hipster).


Now please tell me?

What could you live without?

 

I cannot live without my two furry babies, so a flat where they are welcomed is always a must.




If you found the post interesting please subscribe, comment and share it! Your interaction is crucial to make this site grow up!

 >> MACHINE LEARNING FOR BEARDED HIPSTERS SERIES <<


Download the solving problems page of the Machine Learning for Bearded Hipsters. It is completely FREE!

>> DOWNLOAD HERE <<


 

xoxo,

3 Replies to “Solving Machine Learning Problems – Falling in reverse without realizing it.”

  1. My brother suggested I might like this web site. He was totally right.
    This post actually made my day. You cann’t imagine just how much time I had spent for this info!
    Thanks!

Leave a Reply

Your email address will not be published.