[...] Basics
What is artificial intelligence? Machines that mimic human behaviour
What is machine learning? Computer algorithm improving thru data
What is deep learning? Machines w/ artificial neural network to solve complex problems e.g. more than 2 hidden layers
Key elements of artificial intelligence:
- Machine learning Machines that mimic human behaviour
- Anomaly detection Detect outliers and things out of place
- Computer vision See and identify things as a human would
- Natural language processing Process human language and infer context
- Conversational AI Hold a conversation w/ a human
What is data labelling? Adding meaningful labels to provide context on identified raw data
What is supervised learning? Make prediction using labelled data e.g. classification, regression
What is unsupervised learning? Algorithm labels the unlabelled data by pattern recognition e.g. clustering, dimensional reduction, association
What is reinforcement learning? Uses algorithm that learns from goal and take steps needed e.g, game AI, robot navigation
What is neural network? Neuron represents an algorithm, mimicking the brain
6 principles for responsible AI:
- Fairness Everyone should be treated fairly
- Reliability & safety
- Privacy & security
- Inclusiveness AI systems designed to empower and engage individuals
Users should be made fully aware of the purpose of the system, how it
- Transparency works, and what limitations may be expected.
- Accountability People to be held accountable for AI systems
How to predict numerical value? Regression e.g. how much will the house cost next year
How to predict label as category? Classification e.g. predict the type of flower for given petal lengths and color, yes/no
How to divide data into groups? Clustering e.g. which viewers like the same type of movies
What algorithm does logistic regression belong to? Classification
Clean collected data, train the model and compute it before
Describe the general process of AI registering the model. Deploy the completed model using API
What are the operations in azure ml called? Experiments. Auto ml finds the best performing model for data
What does a low normalized root mean square error mean? Smaller value indicates higher accuracy of the model prediction
Which deployment (ACI / AKS) for production? Azure kubernetes service * ACI = azure container instances
What is the name of Azure's computer vision service? Computer vision cognitive service
What information does computer vision resources provide in Azure?
A key and an endpoint
A machine learning algorithm to predict which category an image e.g. product identification, disaster investigation, medical
What is image classification? belongs to diagnosis
Describe the classification machine learning model process Set of inputs are used to calculate a probability score and predict its label
What is the purpose of convolutional neural networks (CNN)? Process pixels and uncover the patterns within it Deep learning
What is artificial intelligence? Machines that mimic human behaviour
What is machine learning? Computer algorithm improving thru data
What is deep learning? Machines w/ artificial neural network to solve complex problems e.g. more than 2 hidden layers
Key elements of artificial intelligence:
- Machine learning Machines that mimic human behaviour
- Anomaly detection Detect outliers and things out of place
- Computer vision See and identify things as a human would
- Natural language processing Process human language and infer context
- Conversational AI Hold a conversation w/ a human
What is data labelling? Adding meaningful labels to provide context on identified raw data
What is supervised learning? Make prediction using labelled data e.g. classification, regression
What is unsupervised learning? Algorithm labels the unlabelled data by pattern recognition e.g. clustering, dimensional reduction, association
What is reinforcement learning? Uses algorithm that learns from goal and take steps needed e.g, game AI, robot navigation
What is neural network? Neuron represents an algorithm, mimicking the brain
6 principles for responsible AI:
- Fairness Everyone should be treated fairly
- Reliability & safety
- Privacy & security
- Inclusiveness AI systems designed to empower and engage individuals
Users should be made fully aware of the purpose of the system, how it
- Transparency works, and what limitations may be expected.
- Accountability People to be held accountable for AI systems
How to predict numerical value? Regression e.g. how much will the house cost next year
How to predict label as category? Classification e.g. predict the type of flower for given petal lengths and color, yes/no
How to divide data into groups? Clustering e.g. which viewers like the same type of movies
What algorithm does logistic regression belong to? Classification
Clean collected data, train the model and compute it before
Describe the general process of AI registering the model. Deploy the completed model using API
What are the operations in azure ml called? Experiments. Auto ml finds the best performing model for data
What does a low normalized root mean square error mean? Smaller value indicates higher accuracy of the model prediction
Which deployment (ACI / AKS) for production? Azure kubernetes service * ACI = azure container instances
What is the name of Azure's computer vision service? Computer vision cognitive service
What information does computer vision resources provide in Azure?
A key and an endpoint
A machine learning algorithm to predict which category an image e.g. product identification, disaster investigation, medical
What is image classification? belongs to diagnosis
Describe the classification machine learning model process Set of inputs are used to calculate a probability score and predict its label
What is the purpose of convolutional neural networks (CNN)? Process pixels and uncover the patterns within it Deep learning
