Everything you need to know about Machine Learning Contest

What is Machine Learning? 

Machine Learning is the science of using data to teach computers how to mimic intelligent human behaviours. It refers to anything from programs that teach computers how to play board games and translate languages to complex systems which can have human-sounding conversations or help to diagnose health conditions. Check out this video to understand more about Machine Learning: What is Machine Learning?

Can you give me examples of Machine learning from real life? 

There is Machine Learning technology all around us. Machine Learning has been used for things like recommending you the next video based on your likes and recently to help predict floods in India. Machine learning can also help us protect our whales by listening to whale songs.  

Why do we need Machine Learning? Why can’t we just learn ourselves? 

On an everyday basis, there is a lot of data that we consume and create. Think of having a big library of photos and your friend asks you - “hey, show me the photos from your last vacation to Shimla”. Imagine if you have 100 photos on your phone out of which 50 are from the vacation -  then you will be able to find the photos for your friend in some time. 

Now think if you had 1500 photos on your phone, then it will be very tiring and time consuming for you to find those photos. 

Let’s think of an alternate solution - what if you could say to your Google Photos app “show me my photos from Shimla” and it shows you them without any effort. Voila! 

This is the benefit of machine learning - it helps people and companies make quicker and accurate decisions. 

Do I have to write programs to make machines learn for this contest? 

No, absolutely not. You will be using the tool developed by Google Engineers called Cloud AutoML. With AutoML, you will not need to create and code models because they are already created for you. So get your thinking cap on and think about what will be a good problem to work on using these already ready machine learning models. 

What are the kind of models you have? 

We have two products most relevant to you that will help you with two unique traits, we advise you to pick any one (we suggest to start with Auto ML Vision but you can use what you like) 

  1. AutoML Vision: Derive insights from images in the cloud

  2. AutoML Natural Language: Reveal the structure and meaning of text through machine learning 

What data do I have to give to AutoML Products to train it? Also, what is training? 

Data is important for the Machine to learn. So if you are using AutoML Vision then you need to give it images and tell it what those images are (by labelling the image) to understand and learn that for future prediction. If you are using AutoML Natural Language then you need to give text data and tell the computer what that text is. 

Think of it this way, you upload data on AutoML and tell the computer what that data is (called labelling) so it can learn that information for future prediction. 

Let us understand this with an example: 

  • Identifying the Problem: You notice that people around you put waste always in the wrong dustbin. So you decide to create an application where they click a picture of their waste and the app tells whether it goes in the Wet Waste dustbin or Dry Waste Bin. So you start by clicking pictures of waste and want the computer to identify whether it is wet waste (food waste) or dry waste (plastic, paper, etc)

  • Collecting and Labelling the Data: You click pictures of (for eg.) 100 different pictures of waste (banana peels, pizza, plastic bottles, paper napkin etc) and label 50 as ‘dry waste’ and 50 as ‘wet waste’

  • Training the Model: You then use AutoML Vision Train Model so that the computer can learn what is wet waste and dry waste 

  • Testing the Model: Then you try out your model by checking if the computer can predict an image that it has never seen before (that it was not trained on) and see how the computer identifies it. 

What do I have to do for the Code to Learn Contest? 

We will NOT be expecting you to create your own models, instead we will be assessing you on 

  • How you define a problem statement: Define the Problem you want to work on. For instance, in the previous example, we defined a problem to identify waste segregation as a use of machine learning. It can have various uses, it can help someone if you are confused what dust bin to put the waste in then capture an image and give it  the model and it will tell you the right one 

  • How and what Data you choose: Identify a data set that you want to work on. You can also create your own data set. Just like we did in the previous example by capturing images 

    • There are a lot of resources you can find on the web for data sets. Searching for keywords like “Open Data Sets for Natural Language Processing” or “Open Data sets for Vision” will help. Try your own search queries :) 

What is Google Cloud and what is got to do with Cloud AutoML? 

Google Cloud is a platform to perform computing services, it is the same infrastructure Google uses to run its Search Platform. 

Cloud AutoML (https://cloud.google.com/automl/) is a product on Google Cloud to do Machine Learning without creating your own models. 

How to use Google Cloud? Will I be charged for it? 

  • Create a Google Cloud Account using your Gmail account and get started with your free trial. Unfortunately, you will not be able to create a Google cloud AutoML Project without a Gmail account

  • Please do NOT create a Google Cloud account without parental consent

  • It will ask you for Credit Card details but you will NOT be charged on your card once you complete the free credits (unless you manually use that option) 

  • All new users get USD 300 worth credits

  • If you have exhausted your free credits with the existing account, we have the option of availing additional credits through vouchers. You can reach out to us at codetolearn@google.com for a voucher after considering the following points:

    • It would not be possible for you to redeem these vouchers to pay your bills after you have spent an amount on the Google Cloud Platform. Therefore, you should reach out to us in advance before starting the Code to Learn project

    • Only those users who have exhausted their free trial credits are eligible for a voucher

    • Please request for vouchers only after taking parental consent

    • Vouchers are only available till 25th June, 2020

    • Note: Vouchers can only be redeemed for creating Code to Learn project

  • Read the FAQs on free trial here

What format/file do I need to submit as a part of the Code to Learn contest? 

Please refer to the project submission page for submission guidelines.

Sounds great, can you point me to some learning resources? 

Well, there are a lot of resources to use AutoML on the open web. For eg: try searching for learning videos on YouTube. You can also check out the materials available here to get started!