Machine Learning and AI on AWS: Complete Guide

Discover the AWS services for developing and deploying Artificial Intelligence solutions.

1. Amazon SageMaker

Environment Setup

# SageMaker configuration example
import boto3
import sagemaker

def setup_sagemaker():
    sagemaker_session = sagemaker.Session()
    
    # Create the notebook instance
    notebook_instance = sagemaker_session.create_notebook_instance(
        NotebookInstanceName='MonNotebook',
        InstanceType='ml.t3.medium',
        RoleArn='arn:aws:iam::account:role/AmazonSageMaker-ExecutionRole',
        SubnetId='subnet-xxxxx',
        SecurityGroupIds=['sg-xxxxx']
    )
    
    return notebook_instance['NotebookInstanceArn']

Training a Model

# Example of model training with SageMaker
def train_model():
    sagemaker_session = sagemaker.Session()
    
    # Training configuration
    estimator = sagemaker.estimator.Estimator(
        image_uri='763104351884.dkr.ecr.region.amazonaws.com/pytorch-training:1.8.1-cpu-py3',
        role='arn:aws:iam::account:role/AmazonSageMaker-ExecutionRole',
        instance_count=1,
        instance_type='ml.c5.xlarge',
        volume_size=30,
        output_path='s3://mon-bucket/output/',
        sagemaker_session=sagemaker_session
    )
    
    # Start training
    estimator.fit({
        'train': 's3://mon-bucket/data/train/',
        'validation': 's3://mon-bucket/data/validation/'
    })
    
    return estimator.latest_training_job.name

2. Amazon Rekognition

Image Analysis

# Example of image analysis with Rekognition
def analyze_image():
    rekognition = boto3.client('rekognition')
    
    # Analyze an image
    response = rekognition.detect_labels(
        Image={
            'S3Object': {
                'Bucket': 'mon-bucket',
                'Name': 'mon-image.jpg'
            }
        }
    )
    
    # Face detection
    faces = rekognition.detect_faces(
        Image={
            'S3Object': {
                'Bucket': 'mon-bucket',
                'Name': 'mon-image.jpg'
            }
        },
        Attributes=['ALL']
    )
    
    return {
        'labels': response['Labels'],
        'faces': faces['FaceDetails']
    }

Text Detection

# Example of text detection with Rekognition
def detect_text():
    rekognition = boto3.client('rekognition')
    
    # Detect text in an image
    response = rekognition.detect_text(
        Image={
            'S3Object': {
                'Bucket': 'mon-bucket',
                'Name': 'mon-document.jpg'
            }
        }
    )
    
    return response['TextDetections']

3. Amazon Comprehend

Text Analysis

# Example of text analysis with Comprehend
def analyze_text():
    comprehend = boto3.client('comprehend')
    
    # Sentiment analysis
    sentiment = comprehend.detect_sentiment(
        Text='Ce produit est excellent !',
        LanguageCode='fr'
    )
    
    # Entity detection
    entities = comprehend.detect_entities(
        Text='Amazon Web Services est une filiale d\'Amazon.',
        LanguageCode='fr'
    )
    
    return {
        'sentiment': sentiment['Sentiment'],
        'entities': entities['Entities']
    }

Document Classification

# Example of classification with Comprehend
def classify_document():
    comprehend = boto3.client('comprehend')
    
    # Document classification
    response = comprehend.classify_document(
        Text='Ceci est un document technique concernant l\'intelligence artificielle.',
        EndpointArn='arn:aws:comprehend:region:account:document-classifier/mon-classifier'
    )
    
    return response['Classes']

4. Amazon Polly

Speech Synthesis

# Example of speech synthesis with Polly
def synthesize_speech():
    polly = boto3.client('polly')
    
    # Convert text to speech
    response = polly.synthesize_speech(
        Text='Bonjour, je suis un assistant virtuel.',
        OutputFormat='mp3',
        VoiceId='Lea',
        LanguageCode='fr-FR'
    )
    
    # Save the audio file
    with open('output.mp3', 'wb') as file:
        file.write(response['AudioStream'].read())
    
    return 'output.mp3'

Lexicon Management

# Example of lexicon management with Polly
def manage_lexicon():
    polly = boto3.client('polly')
    
    # Create a custom lexicon
    lexicon = {
        'lexiconName': 'MonLexique',
        'content': json.dumps({
            'lexicon': [
                {
                    'grapheme': 'AWS',
                    'phoneme': 'A W S'
                }
            ]
        })
    }
    
    response = polly.put_lexicon(**lexicon)
    return response['Lexicon']['Name']

5. Amazon Personalize

Service Configuration

# Personalize configuration example
def setup_personalize():
    personalize = boto3.client('personalize')
    
    # Create a dataset group
    dataset_group = personalize.create_dataset_group(
        name='MonDatasetGroup',
        roleArn='arn:aws:iam::account:role/service-role/AmazonPersonalize-ExecutionRole'
    )
    
    # Create a schema
    schema = personalize.create_schema(
        name='MonSchema',
        schema='{"type": "record", "name": "Interactions", "namespace": "com.amazonaws.personalize.schema", "fields": [{"name": "USER_ID", "type": "string"}, {"name": "ITEM_ID", "type": "string"}, {"name": "EVENT_TYPE", "type": "string"}, {"name": "TIMESTAMP", "type": "long"}]}'
    )
    
    return {
        'dataset_group_arn': dataset_group['datasetGroupArn'],
        'schema_arn': schema['schemaArn']
    }

Creating Recommendations

# Example of creating recommendations
def create_recommendations():
    personalize = boto3.client('personalize')
    
    # Create a solution
    solution = personalize.create_solution(
        name='MonSolution',
        datasetGroupArn='arn:aws:personalize:region:account:dataset-group/MonDatasetGroup',
        recipeArn='arn:aws:personalize:::recipe/aws-user-personalization'
    )
    
    # Create a campaign
    campaign = personalize.create_campaign(
        name='MonCampaign',
        solutionVersionArn=solution['solutionArn'],
        minProvisionedTPS=1
    )
    
    return {
        'solution_arn': solution['solutionArn'],
        'campaign_arn': campaign['campaignArn']
    }

Conclusion

Key takeaways:

• Develop ML models with SageMaker
• Analyze images with Rekognition
• Process text with Comprehend
• Generate speech with Polly
• Create recommendations with Personalize

Recommendations:

• Choose the right services for your needs
• Optimize processing costs
• Set up effective monitoring
• Secure your data
• Follow AWS best practices