AWS Machine Learning Certification

The ‘AWS Machine Learning – Specialty’ certification is tough. It is designed to test 1-2 years of practical machine learning knowledge, plus AWS machine learning implementation specifics. It’s broad – you’ll also need AWS systems architect knowledge to get through this one.

AWS’s own Machine Learning certification path comes with some useful questions and a full test – https://aws.amazon.com/training/learning-paths/machine-learning/exam-preparation, but the course itself isn’t ideal.

The paid AWS practice exam has only a handful of questions. So, over the next few weeks I’m providing some original and free practice questions. These should make you think, and help you check for any knowledge gaps. Questions are aimed to be at around the same level, most are in the same format. However, in some cases they may require more specific knowledge than might be expected in the exam.

Enjoy, and good luck!

256
Created on By adminadmin

AWS Machine Learning – Quiz 1

1 / 10

QUESTION

A data scientist has built a multi-class burger classifier, with the aim to identify the corresponding meal from social media images.

Given the resulting multi-class confusion matrix from the first training run, what are the missing values x, y, z in the metrics table that follows?

PREDICTED
big mac meal
whopper meal
in & out        double-double
ACTUAL
big mac meal
5
5
10
whopper meal
5
10
5
in & out double-double
10
5
5

 

 

precisionrecallf1-scoresupport
big mac mealx0.250.2520
whopper meal0.5y0.520
in & out double-double0.250.25z20

(Select One)

2 / 10

QUESTION

A national tourist agency data team are using Blazing Text to classify social media posts to specific activities types and overall positive or negative classification. Which of the following are true?

a) For training, raw text must be converted to space separated tokenised text.For pipe mode, there is no need to use RecordIO. Both train and validation channels are supported
b) The algorithm supports only binary classification
c) Only CPU is supported
d) The algorithm extends the FastText classifier and is an unsupervised learning model.
e) Accuracy on the validation data is used as a proxy to the quality of the algorithm

 

(Choose One)

3 / 10

QUESTION

A data scientist is working with a small dataset of 1000 rows and 6 features (labelled f1, f2 … f6)
Feature F2 is numerical and has about 80 entries with no value set. The data is going to be used to create a simple regression model.

How should the missing values be handled?

4 / 10

QUESTION

Which of these would provide the most manageable, scalable and secure means for an app to access inference from a SageMaker Endpoint?

(Choose One)

5 / 10

QUESTION

Which of the following must be set by the user (required hyperparameters), for SageMaker’s built-in algorithm, DeepAR Forecasting?

(Select Three)

6 / 10

QUESTION

A data scientist is working on a machine learning model to improve relevance in search results. As a benchmark, the scientist is  using SKLearn’s TF-IDF implementation.

However, at first glance, the td-idf values aren’t looking as expected. Which of the following best explains why the values might different from the scientists initial calculations? And, given the following corpus, what are the dimensions for the tf-idf matrix if only bigrams are selected.

Document1 : product managers know about machine learning
Document2: machine learning, a product owner essential

(Choose One)

 

7 / 10

QUESTION

A data scientist wishes to create word embeddings for use in downstream modelling and has selected Blazing text for evaluation.
Which of the following are true?

(Choose One)

 

8 / 10

QUESTION

A data scientist is evaluating a machine learning model – which of the following are true?

(Select two)

9 / 10

QUESTION

Which of the following must be set by the user (required hyper parameters), for SageMaker’s built-in algorithm, K-Nearest Neighbours?

(Select Three)

10 / 10

QUESTION

Which of the following SageMaker algorithms or AWS Services offer access to the benefits of transfer learning?

a) Image Classification Algorithm
b) Object Detection Algorithm
c)Semantic Segmentation Algorithm
d) Amazon Rekognition

 

(Choose One)

Your score is

The average score is 46%

0%

3 Responses

    • For this question……

      A data scientist is working with a small dataset of 1000 rows and 6 features (labelled f1, f2 … f6)
      Feature F2 is categorical and has about 80 entries with no value set. The data is going to be used to create a simple regression model.

      How should the missing values be handled?

      Drop rows with missing data
      Use the mean for this feature for missing values
      Use KNN to determine average values to replace missing value.
      Use a deep learning to impute the missing values

  1. Hello Sanket,

    Many thanks for taking the time to provide feedback. Yes, you’re right and I’ve corrected the question.

Leave a Reply

Your email address will not be published. Required fields are marked *