CHAPTER: ARTIFIICAL INTELLIGENCE

Question 1

Artificial Intelligence

Answer 1

• the ability of a computer to perform tasks that usually only humans can do
  like decision-making, speech recognition

Question 2

Machine Learning

Answer 2

• subset of AI
• computers learn without explicit programming
• ML computers fed with old training data, producing models from which predictions about previously unseen data can be made

Question 3

Deep Learning

Answer 3

• subset of ML
• where computers learni to solve problems using neural networks similar to how the human brain functions
• makes use of artificial neural networks to extract patterns from data

Question 4

TYPES OF ML: SUPERVISED LEARNING

Answer 4

• learns using labelled data
• using known tasks with given outcomes to enable a computer program to improve its performance in accomplishing similar tasks
• common problem types : regression and classification
• aim : calculate outcomes
• eg: risk evaluation , forecast sales

Question 5

TYPES OF ML: UNSUPERVISED LEARNING

Answer 5

• learns by using unlabelled data
• using large no of tasks with unknown outcomes to enable computer program to improve performance in accomplishing tasks
• common problem types : association and clustering
• aim : discover underlying patterns
• eg: recommendation system , anamoly detection

Question 6

TYPES OF ML: REINFORCEMENT LEARNING

Answer 6

• works on interacting with environment
• using large number of tasks with unknown outcomes + the feedback to enable computer program to improve its performance in accomplishing tasks
• common problem type : exploitation and exploration
• aim : learn series of actions
• eg: self driving cars , healthcare

Question 7

Artificial Neural Networks

Answer 7

• artificial network built from networks
• group of interconnected input and output units where connection (neuron) is either activated or not

Question 8

LAYERS OF ANN: Input

Answer 8

• accepts several different forms of input

Question 9

LAYERS OF ANN: Hidden Layer

Answer 9

• present in between the I/O layers
• performs all calculations to find hidden features and patterns

Question 10

LAYERS OF ANN: Output Layer

Answer 10

• output conveyed
• calculate value must meet value threshold to be output

Question 11

Purpose of many hidden layers

Answer 11

• allows for deep learning to take place
• higher complexity = higher number of layers used to solve the problem
• allows Neural Networks to learn and make decisions independently
• improve the accuracy of results

Question 12

How does ANN enable ML

Answer 12

• ANN replicates the way of the human brain
• there is weightings allocated for each connection between each node
• data is inputted in the input later and outputs results in the output layer
• input is analysed at reach hidden layer to calculate the outputs
• the training process is repeated to ensure the get bes possible output (reinforcement learning)
• back propagation of errors to fix any mistakes made

Question 13

The Process of BackPropogation

Answer 13

• initial output compared to the expected value
• difference calculated
• outputs travel to hidden later to adjust weightings in each neuron
• this si sot minimise the difference
• iterative process until acceptable error change or no change at all

Question 14

Explain the use of graphs and give examples where they are used

Answer 14

• provide relationship between nodes
• AI problems solved by finding path in graph
• graphs can be analysed by range of algorithms

eg:
used to represent ANN
tells relationship between nodes
used in ML