Make your life easier with AutoML Tables on Google Cloud!

Elias Papachristos
Oct 29 · 5 min read

We are living in the era of the Cloud. And that’s a fact!

Not only we can do things faster but also cheaper!

For everyone who has worked with AI-ML-DL models on the Cloud, training-evaluation, can easily understand what I’m talking about. You can check your model, change hyperparameters… and get results very fast, comparing doing them on-prem. Our time is precious!

I wrote that it’s cheaper. It’s true. We can try many different models at the same time than waiting for the results of the 1 model on-prem. Time Is Money after all!

The main reason that I’m writing this article is the speed of working on the Cloud. I will talk for Google Cloud, the platform and the AutoML Tables (still in Beta, at least until today 29/10/2019).

The weekend 11–13/10/2019 was the NASA Space Apps Challenge 2019, in Thessaloniki and around the world.

NASA Space Apps Challenge 2019 Thessaloniki

I was there as a Mentor, awesome experience by the way! While I was moving from one team to another, on Saturday evening, I came up with a problem. The programmer of one team didn’t have time to write the code, train-evaluate the model and make the prediction.

So, I suggested him to try the Auto ML Tables. In a few words, he had to sign-up on Google Cloud Platform, import the data, click the button to start the analysis, click the button to start the training and get the evaluation. Get the scores and then deploy the model for real-time predictions (if he wanted to). He did it and had the results ready the next morning!

If you read carefully the above paragraph you will see that I’m not writing anything about coding. Yes, it’s true! You can have your model up and running without coding.

So, let’s see how it works.

First of all, go to the Google Cloud Platform and sign-in or sign-up. You will get $300 or one year of use for free (whatever ends first). Create your project and connect it with your billing account (you can read more here

Before using the AutoML Tables it’s better to clean your data and store the data either on Cloud Storage or on BigQuery.

I prefer to store the data on Cloud Storage (create bucket) and then import it on BigQuery, at the dataset that I’ve created. By importing my CSV file from Cloud Storage to BigQuery I create my table at the same time. There I can make my last changes with SQL.

Everything is ready to import my data to the AutoML tables. If you scroll down at the Google Cloud Platform you will see Tables, at the PRODUCTS section.

Tables page

By clicking the New Dataset a pop up will appear and you will give your dataset a name. Then you have to choose if you will import your data from BigQuery, Cloud Storage or from your computer. Check it and fill the inputs. Then press Import.

Tables’ Schema

After a while, the schema of your dataset will appear. The main thing to remember is to give your Target column. If you agree with the schema you can press the Continue button. And the Analysis will begin.

Tables’ Analysis

Until this stage, you can go back, make changes to your data, import it and analyze it again without paying anything. If you are ready then you can click on the TRAIN.

Before starting the Train you must give how many hours you want it to be trained. You can train a model for up to 6 hours for free. You get another $100 for the AutoML Tables.

Tables’ Train

I decided to train my model for 3 hours. It was “working” for 4. But, I wasn’t charged for the infrastructure that was built for me. Only for the 3 hours (~$50) that I applied.

The results of my Binary Classification model were great! And all of these without writing any line of code! Ok, that’s not exactly true. I had to write code to clean the data but I didn’t do any coding for the algorithm, hyperparameters… and all the rest.

Tables’ Evaluate

Three Main Points to remember about AutoML Tables:

AutoML Tables is a supervised learning service. This means that you train a machine learning model with example data. AutoML Tables uses tabular (structured) data to train a machine learning model to make predictions on new data.

..there are recommended minimum rows of example data:

Classification problem: 50 x the number features

Regression problem: 200 x the number of features

Free trial: You can try AutoML Tables for free by using 6 free node hours each for training and for batch prediction, per billing account. Your free node hours are issued right before you create your first model, and you have up to one year to use them.

You can read more in details here:

What I loved is that while importing my data and training my model I could do something else. Because at each one of these stages I was informed by email the time it finished.

For sure, it’s another great technology that makes our life easier. And it would be better if, after the evaluation, we could have the algorithm that was used with all the details that gave as these results. Don’t you agree?

Elias Papachristos

Written by

Full-Time Family Man, Ex-Military Helicopter Pilot, Kendo Practitioner, Developer, Lead GDG Cloud Greece, Beta Tester Coursera(Volunteer), AI Enthusiast!

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade