AI, ML, and Deep Learning
AI, ML, and DL Overview
Understanding AI, ML, DL
Initial Demo Script
Demo Script - Introduction
1AI - Artificial Intelligence
2ML - Machine Learning
3DL - Deep Learning
4
5Amazon SageMaker
6Amazon Rekognition
7
8Recommendations, Predictions and Forecasting AI and ML is necessary.
9
10AI:
11--->
121. Visual Perception
132. Speech recognition
143. Decision-making
154. Language translation
16
17AI ----> ML(Machine Learning), DL(Deep Learning)
18
19AI ---> Narrow AI / Generative AI
20
21ML: Train models to predict
22DL: Deep learning is a subset of ML that uses multi-layered neural-networks to model and solve more complex problems.
23
24
25
26AI - General Intelligence Simulations
27ML - Learning from data patterns
28DL - Complex pattern recognition with neural networks
Inferencing Concepts
Types of Inferencing
Inferencing Types
1Inferencing is the process in which AI models can make predictions using data
2
3Real-time inferencing
4Batch inferencing
Text Data and NLP Models
Data Types in AI Models
Data Types in AI
1Numerical
2Categorical
3text
4images
5
6
7AWS Glue
8AWS Elastic Map Reduce
9Data wrangler
10
11Numerical Data:
12
13Amazon Sagamaker able to load numerical data-sets from s3, redshift etc...
14
15
16Categorical Data:
17
18Text Data and Natural Language Processing(NLP)
19
20Raw Text Data --> Tokenization, Stop-Word Removal(Pre-processing) ---> NLP model ----> Sentiment Analysis and Text Classification (Output)
21
22
23Image Data:
24
25Image Recognition, Object Detection and Facial Recognition
26Amazon Rekognition
27
28
29Audio Data:
30Speech Rekognition model
31
32Labeled Data vs Unlabeled Data:
33-------------------------------------->
34Labeled data contains both the input data and the corresponding output.
35
36unlaled data has the only input without the expected output.
37
38Time-series data
39Long shirt-term memory networks(LSTM)
40Prophet(a tool in aws)
41
42
43Balanced and imbalanced data.
44
45
46BigData and AI
47
48EMR, AWS Glue
49
50
51BigData -- Apache Hadoop -- AWS EMP(ElasticMapReduce)
52
53Types of Machine Learning:
541. Supervised learning
552. Unsupervised learning
563. Reinforcement Learning
AWS ML Data Collection Tools
AWS Tools for Machine Learning
IaC and MLOps Integration
ML Pipeline Script
ML Pipeline Automation
1ML Model Lifecycle:
2---------------------->
31. Data Collection
42. Model Training
53. Model Deployment
64. Monitoring
7
8
9Hyper parameter tuning
10
11Amazon SageMaker have AMT(Automatic Model Training)
12
13Auto-optimize models by running multiple training jobs.
14
15Model Performance:
16----------------------->
171. Accuracy
182. Precision
193. Recall
204. F1 Score
21
22
23Model Deployment:
24------------------------->
25Real-time
26Batch
27
28Deploy using AWS SageMaker
29
30
31Model Monitor:
32------------------------->
33
34
35Apache Airflow
36
37MLOps Pipeline
38
39
40Amazon SageMaker Clarify
41Amazon Sagemaker studio
42
43
44Model Metrics vs Business Requirements
Amazon Polly Demo
AWS Glue for Data Processing
AWS AI/ML Services Script
AWS Services for AI/ML
1AWS SageMaker - Simplify Machine Learning Development
2 1. Data Preperation
3 2. Model Training
4 3. Hyper parameter Tuning
5
6 Data Stored in S3 - AWS Glue for ETL
7
8
9AWS Polly - Pre-Trained model to enable quick ai integration
10AWS Lex - Pre-trained model to enable quick AI integration
11
12AWS Bedrock - premium GenAI service - Hosts AI models
13Guard Rails for Bedrock
14AWS Bedrock Agents
15
16
17Amazon Queue
18
19
20Amazon Rekognition: It analyzes images and videos for object detection, facial analysis and text recognition
21
22
23Amazon Lex: service for conversational AI - Amazon Alexa
24
25Amazon Polly - Audio interactions.. basically voice based system ...conversion for text to speech API
26SSML - Speech Synthetics Markup Language(SSML)
27
28
29AWS Comprehend:
30 1. Language detection
31 2. Sentiment Analysis
32 3. Entity recognition
33 4. Key phrases extracton
34 5. Topic Modeling
35
36AMazon Fraud detector
37
38Amazon Transcribe.... audio and video to text... i.e. sound into text
39
40Amazon Textract --- extract text from scanned document
41
42
43AWS Glue: Data manipulation and data cleaning and data extraction - Run ETL jobs
44AWS Glue Databrew - Visual Data Preperation tool
45
46AWS Crawler structure the data from s3... json..csv to etc
47
48
49Amazon Elastic MapReduce(haddop cluster, apache spark cluster)
50
51
52Augmented AI - Make predictions in human reviews
53
54QuickSight- A visualization Engine -- uses SPICE Engine