Your proven path to success in Machine Learning Interviews – developed by FAANG engineers. Unlock ML loops at top companies with a System Design approach.

System design is an important component of any ML interview. Being able to efficiently solve open-ended machine learning problems is a key skill that can set you apart from other engineers and increase the level of seniority at which you’re hired.

This course helps you build that skill, and goes over some of the most popularly asked interview problems at big tech companies. You’ll walk step-by-step through solving these problems, focusing in particular on how to design machine learning systems rather than just answering trivia-style questions. 

Once you’re done with the course, you’ll be able to not just ace the machine learning interview at any tech company, but impress them with your ability to think about systems at a high level. If you have a machine learning or system design interview coming up, you’ll find the course tremendously valuable.

Grokking the Machine Learning Interview

The performance of the user engagement prediction model will depend drastically on the quality and quantity of training data. So let's see how the training data for our model can be generated.




# Training data generation through online user engagement

When we show an ad to the user, they can engage with it or ignore it. 
Positive examples result from users engaging with ads, e.g., clicking or adding an item to their cart. Negative examples result from users ignoring the ads or providing negative feedback on the ad.




Suppose the advertiser specifies "click" to be counted as a positive action on the ad. In this scenario,  a user-click on an ad is considered as a positive training example, and a user ignoring the ad is considered as a negative example.

Suppose the ad refers to an online shopping platform and the advertiser specifies the action “add to cart” to be counted as positive user engagement. Here, if the user clicks to view the ad and does not add items to the cart, it is counted as a negative training example.

# Balancing positive and negative training examples

Users' engagement with an ad can be fairly low based on the platform e.g. in case of a feed system where people generally browse content and engage with minimal content, it can be as low as 2-3%.

How would this percentage affect the ratio of positive and negative examples on a larger scale?


Let's look at an extreme example by assuming that one-hundred million ads are viewed collectively by the users in a day with a 2% engagement rate. This will result in roughly $two$ million positive examples (where people engage with the ad) and $98$ million negative examples (where people ignore the ad).

In order to balance the ratio of positive and negative training samples, we can **randomly down sample** the **negative examples** so that we have a similar number of positive and negative examples.


Let's learn about the techniques for generating training data for the ad prediction system.

Introduction

Practical ML Techniques/Concepts

Search Ranking

Feed Based System

Recommendation System

Self-Driving Car: Image Segmentation

Entity Linking System

Ad Prediction System

Training Data Generation

Training data generation through online user engagement

Balancing positive and negative training examples