Sagemaker Built-In Algorithms

Question 1

XGBoost

Answer 1

• supervised learning
• an implementation of the gradient-boosted trees algorithm
• combines an ensemble of estimates from a set of simpler and weaker models.
• highly effective for structured/tabular data
• applicable for both classification and regression

Question 2

Linear Learner

Answer 2

• supervised learning
• used for binary and multi-class classification as well as linear regression

Question 3

Factorization Machines

Answer 3

• supervised learning
• high-dimensional sparse datasets
• ideal for recommendation systems and click prediction
• good at capturing interactions between features

Question 4

K-Means

Answer 4

• unsupervised learning
• For clustering data points into groups based on similarity.

Question 5

Random Cut Forest

Answer 5

• unsupervised
• designed for anomaly detection in large datasets.

Question 6

Image Classification

Answer 6

• Computer vision
• Based on ResNet architecture for categorizing entire images.

Question 7

Object Detection

Answer 7

• Computer vision
• Utilizes Single Shot Detector (SSD) to identify and locate multiple objects within an image.

Question 8

Semantic Segmentation

Answer 8

• Computer vision
• Employs the Fully Convolutional Network (FCN) algorithm for pixel-level image classification, providing detailed object shapes.
• segmentation is a pixel-wise classification problem.
• Attempt to predict a label for every pixel in the image. Results in a mask for which parts of the image correspond to a given label.

Question 9

BlazingText

Answer 9

• NLP
• Supervised text classification (can’t discover topics)
• Highly optimized implementations of Word2Vec (embeddings) and text classification

Question 10

LDA

Answer 10

• For discovering topics within text documents.
• NLP
• unsupervised
• Latent Dirichlet Allocation aka Neural Topic Modeling

Question 11

Sequence2Sequence

Answer 11

• NLP
• Summarization and machine translation.

Question 12

DeepAR

Answer 12

• Time series forecasting
• For forecasting future values in time series data.

Question 13

LightGBM

Answer 13

• Light Gradient Boosting Machine
• faster training speed and lower memory usage compared to other gradient boosting frameworks like XGBoost
• good for classification and regression

Question 14

AutoGluon-Tabular

Answer 14

• open-source AutoML library that automates various aspects of ML, such as model selection, hyperparameter tuning, and ensembling, to achieve high accuracy with minimal effort.
• tabular data

Question 14

CatBoost

Answer 14

• gradient-boosting algorithm designed to handle categorical features efficiently
• doesn’t require explicit one-hot encoding

Question 15

TabTransformer

Answer 15

• supervised learning on tabular data
• tabular classification and regression tasks

Question 16

Regression

Answer 16

• a type of supervised learning used to predict continuous numerical values
• aims to find the best-fitting curve (or line in simpler cases) that represents the relationship between independent variables (features) and a dependent (target) variable.

Question 17

Levenshtein distance

Answer 17

• sometimes called ‘edit distance’
• tells you the closeness of two strings. For example, the word “raed” is closer to the word “read” than the word “baed”.
• applications like autocorrect that matches the correct spelling of a given word.

Question 18

How do you enable XGBoost to perform classification tasks?

Answer 18

To enable XGBoost to perform classification tasks, set the objective parameter to multi:softmax and specify the number of classes in the num_class parameter.

Question 19

Support Vector Machines (SVM)

Answer 19

• supervised algorithm mainly used for classification tasks
• It uses decision boundaries to separate groups of data.

Question 20

SVM with Radial Basis Function (RBF) kernel

Answer 20

• a variation of the SVM (linear) used to separate non-linear data.
• Separating randomly distributed data in a two-dimensional space can be a daunting and difficult task.
• The RBF Kernel provides an efficient way of mapping data (e.g., 2-D) into a higher dimension (e.g, 3-D). In doing so, we can conveniently apply the decision surface/hyperplane where we mainly based our model predictions.

Question 21

Multinomial logistic regression

Answer 21

• used to predict categorical placement in or the probability of category membership on a dependent variable based on multiple independent variables
• wants to understand which ice cream flavors are most preferred by different groups of people

Question 22

What input data types are supported?

Answer 22

recordIO
JSON
JPEG
PNG
CSV
libSVM

No

TSV, TFRecord, HDF5

Question 23

Logistic Regression

Answer 23

Logistic Regression is incorrect as this type of regression only predicts a binary output such as “0” or “1”.

Question 24

knn

Answer 24

supervised

Question 25

word2vec

Answer 25

- text classification, sentiment analysis, entity recognition, translation - not for summarization! use seq2seq for that

Question 26

object2vec

Answer 26

- use when you have labelled objects/abstractions (e.g. claims)