AI Learnings in OCI

What is AI
Cognitive Ability and Problem Solving Capabaility --> Human Intelligence

Human Intelligence if replicated in machines --> Artificial General Intelligence(AGI)

AGI with specific and narrow objective --> AI

AI Terminology

ML --> Machine Learning
DL --> Deep Learning
DS --> Data Science

AI Domains

1. Language Translation
2. Speech Recognition
3. Text to Speech
4. Anamoly detection
5. Product Recco
6. Learn by Reward
7. Weather Forecast
8. Image from Text

Language AI Model

Recurrent Neural Network: process data sequentially and store hidden states

Long Short Term Memory: process data sequentially and retain context

Transformers: Process data in parallel. Self attention to better understand the context.

Audio Speech Model

Recurrent Neural Network
Long Short Term Memory
Transformers
Variational Autoencoders
Waveform Models
Siamese Network

Vision Image Model

convolutional neural networks
YOLO, which is You Only Look Once
Generative adversarial networks

Supervised Learning

Learning from Labeled Data
Classify data or make a prediction

Unsupervised Learning

Data has no outcome or label.
Discovering trends
Understand relationships with data sets
Retail Marketing and Sales
Fruits and Vegetable Nutrition

Reinforcement Learning

Trial and Error
Make decisions and choice

Deep Learning
Extracting features and rules from data.
Neural Network; Supervised Learning

Generative AI

Produce content.

---

Data Type

Numerical data: Measurable data
Categorical data: Characteristic(Nominal or Ordinal)
Time series data: Number Sequence
Text Data: Words and Paragraphs

---

Supervised Learning

• Spam Detection: Categorical Output, Binary in nature
• Disease Detection: Categorical Output, Binary in nature
• Sentiment analysis: Categorical Output, Multiclass in nature

• Stock Price Prediction: Continous and quantitive

• Data Access

• Data Preparation

• Modeling

• Validation

• Deployment

• Monitor and Iteration

Regression --> Linear Regression

Independent Feature (Input) --> Dependent Feature (Output)

Metrics to evaluate Regression
Mean Absolute Error
Mean Squared Error
R Squared

Classification --> Logistics Regression

Metrics to Evaluate Classification

Accuracy: Ration of Correct Prediction/Total number of Prediction

Precision: True positive case/All Positive cases

Recall: Actual Positives identified correctly

Unsupervised Learning

Clustering

• Market Segmentation
• Outlier Analysis
• Reccomdetation System

Reinforcement Learning

• Agent
• Environment
• State
• Action
• Policy, comes from learning

https://dashboard.cohere.com/

Notes taken during learning: OCI AI Foundations Oracle University