Built-in Algorithms

Duration: 60 min

SageMaker provides optimized built-in algorithms for common ML tasks. This module covers XGBoost, Linear Learner, Image Classification, and BlazingText with practical examples and performance considerations.

Overview of Built-in Algorithms

SageMaker built-in algorithms are pre-optimized for performance and scalability. They handle distributed training automatically, support spot instances for cost savings, and integrate seamlessly with SageMaker's training infrastructure.

XGBoost for Regression and Classification

from sagemaker.estimator import Estimator
import sagemaker

session = sagemaker.Session()
role = 'arn:aws:iam::123456789012:role/SageMakerRole'
bucket = session.default_bucket()

# XGBoost container URI
xgboost_container = '246618743249.dkr.ecr.us-east-1.amazonaws.com/sagemaker-xgboost:1.5-1'

# Create XGBoost estimator
xgb_estimator = Estimator(
    image_uri=xgboost_container,
    role=role,
    instance_count=1,
    instance_type='ml.m5.xlarge',
    output_path=f's3://{bucket}/xgboost-output',
    sagemaker_session=session
)

# Set hyperparameters
xgb_estimator.set_hyperparameters(
    objective='binary:logistic',
    num_round=100,
    max_depth=5,
    eta=0.2,
    gamma=4,
    min_child_weight=6,
    subsample=0.8
)

# Train the model
xgb_estimator.fit(
    {'training': f's3://{bucket}/train-data.csv'},
    job_name='xgboost-training-job'
)

Linear Learner for Large-Scale Problems

from sagemaker.linear_learner import LinearLearner

linear_learner = LinearLearner(
    role=role,
    instance_count=1,
    instance_type='ml.m5.xlarge',
    output_path=f's3://{bucket}/linear-output',
    sagemaker_session=session
)

# Set hyperparameters
linear_learner.set_hyperparameters(
    feature_dim=100,
    mini_batch_size=32,
    predictor_type='binary_classifier',
    loss='logistic',
    optimizer='adam',
    learning_rate=0.01,
    epochs=10
)

# Train
linear_learner.fit(
    {'training': f's3://{bucket}/train-data.recordio'},
    job_name='linear-learner-job'
)

Image Classification with ResNet

from sagemaker.image_uris import retrieve

# Get Image Classification container
image_uri = retrieve(
    framework='image-classification',
    region='us-east-1',
    version='latest'
)

image_classifier = Estimator(
    image_uri=image_uri,
    role=role,
    instance_count=1,
    instance_type='ml.p3.2xlarge',
    output_path=f's3://{bucket}/image-output',
    sagemaker_session=session
)

# Set hyperparameters
image_classifier.set_hyperparameters(
    num_classes=10,
    num_layers=50,
    image_shape='3,224,224',
    epochs=30,
    learning_rate=0.01,
    batch_size=32,
    optimizer='sgd'
)

# Train
image_classifier.fit(
    {'training': f's3://{bucket}/image-train/'},
    job_name='image-classification-job'
)

BlazingText for NLP Tasks

from sagemaker.blazingtext import BlazingText

blazingtext = BlazingText(
    role=role,
    instance_count=1,
    instance_type='ml.p3.2xlarge',
    output_path=f's3://{bucket}/blazingtext-output',
    sagemaker_session=session
)

# Set hyperparameters for text classification
blazingtext.set_hyperparameters(
    mode='supervised',
    epochs=5,
    learning_rate=0.05,
    word_ngrams=2,
    vector_dim=100,
    batch_size=32
)

# Train
blazingtext.fit(
    {'training': f's3://{bucket}/text-train.txt'},
    job_name='blazingtext-job'
)

Algorithm Selection Guide

{
  "algorithm_selection": {
    "xgboost": {
      "use_cases": ["Tabular data", "Classification", "Regression"],
      "strengths": ["Fast", "Handles missing values", "Feature importance"],
      "input_format": "CSV or LibSVM"
    },
    "linear_learner": {
      "use_cases": ["Large datasets", "Linear relationships"],
      "strengths": ["Scalable", "Fast inference"],
      "input_format": "RecordIO or CSV"
    },
    "image_classification": {
      "use_cases": ["Image recognition", "Object detection"],
      "strengths": ["Pre-trained models", "Transfer learning"],
      "input_format": "RecordIO or image files"
    },
    "blazingtext": {
      "use_cases": ["Text classification", "Word embeddings"],
      "strengths": ["Fast training", "Supports multiple languages"],
      "input_format": "Text files"
    }
  }
}

Deploying Built-in Algorithm Models

# Deploy XGBoost model
predictor = xgb_estimator.deploy(
    initial_instance_count=1,
    instance_type='ml.m5.large',
    endpoint_name='xgboost-endpoint'
)

# Make predictions
import csv
import io

# Prepare test data
test_data = '5.1,3.5,1.4,0.2'

# Invoke endpoint
response = predictor.predict(test_data)
print(f"Prediction: {response}")

Quiz 1

Quiz 2

Quiz 3

Quiz 4

Quiz 5