atest Submission Grade 100%
1.
Question 1
Why do models become more complex?

1 / 1 point

To reduce GPU usage.


> To increase accuracy.


To cut down costs.


To minimize latency.

Correct
Absolutely! We apply more complex model architectures that allow including more features to increase accuracy.

2.
Question 2
What is the difference between optimizing and satisficing metrics?

1 / 1 point

Optimizing metrics assess model complexity while satisficing metrics evaluate operation costs.


> Optimizing metrics measure the model's predictive effectiveness while satisficing metrics specify operational constraints.


Optimizing metrics estimate the speed of the model’s prediction latency while satisficing metrics deal with its precision.

Correct
Nailed it! First, aim to improve the model's predictive power until the infrastructure reaches a specific latency threshold. Then,  assess the results to approve the model or continue working on it.

3.
Question 3
Which of the following are NoSQL solutions for implementing caching and feature lookup? (Select all that apply)

1 / 1 point

Amazon RDS


> Google Cloud Firestore

Correct
Right on! A good choice for millisecond read latencies on slowly changing data where storage scales automatically.


> Google Cloud Memorystore

Correct
That's right! This database is a good choice for achieving sub-millisecond read latencies on a limited amount of quickly changing data retrieved by a few thousand clients.


> Amazon DynamoDB

Correct
Excellent! Amazon DynamoDB is a scalable low-read latency database with an in-memory cache.

4.
Question 4
True Or False: The main advantage of deploying a model in a large data center accessed by a remote call is that you can disregard costs in favor of model complexity.

1 / 1 point

True


> False

Correct
Exactly! For example, Google constantly looks for ways to improve its resource utilization and reduce costs in its applications and data centers.

5.
Question 5
True Or False: As a rule, you should opt for on-device inference whenever possible.

1 / 1 point

> True


False

Correct
Absolutely! Following this general rule enhances the user experience by reducing the app's response time. There are exceptions, though, such as medical diagnosis, in which the model must be as accurate as possible, and latency is not that important.