1
Q
Explain the use of graph to aid AI
A
- Artificial neural network can be represented using a graph
- Graph shows the relationship between nodes
- AI problem can be defined as finding a path in graph
- graph can be analysed by some algorithms
- Including A* & Dijkstra’s algorithm
- Using ML
- Available methods are back of propagation & regression method
2
Q
Deep learning (DL)
A
- specialised from ML
- use artificial neural network
- Modelled on a human brain
- Contains high number of hidden layer
- Use many layers to progressively extract higher level features.
3
Q
How ANN enable ML ?
A
- ANN intended to replicate the way human brain work
- Weights are assigned for each connection between nodes
- Data is the input in the input layer and are passed into system
- Output layer provides the result
- They are analysed at each subsequent layer where characteristic is extracted
- Reinforcement training take place
- Decisions can be made without being programmed
- The DL net will have created complex feature detectors.
- Back propagation will be used to correct any error that have been made
4
Q
Reason multiple hidden layers in an artificial neural network (ANN) ?
A
- Enable DL to take place
- When problem is more complex, more layers is needed to solve it
- Enable neural network to learn and make decisions on its own
- To improve the accuracy of the result
4
Q
Reasons of using DL
A
- make good use of unstructured data
- Outperform other methods if data is large
- Effective at identifying hidden patterns
- Provide a more accurate outcome with higher numbers of hidden layer
- DL system enables machines to process data with non-linear approach
5
Q
Dijkstra’s algorithm
A
- a method of finding shortest path between 2 points on a graph -
Each point on the graph has a node and vertex
6
Q
AI
A
- intelligent machine think and behave like humans
7
Q
Back propagation
A
– method used in artificial neural networks to calculate error gradients so that actual
node weightings can be adjusted to improve the performance of the model.
8
Q
ML
A
- systems learn without being programmed to learn, subset of AI
- the algorithms are ‘trained’ to make predictions based on previous scenarios.
9
Q
Type of ML
Supervised learning
A
- make use of regression analysis and classification analysis
- Require both input and output to train model
- Predict future outcome based on past data
- Classify labelled data
10
Q
Type of ML
unsupervised learning
A
- Make use of density estimation and k-mean clustering
- able to identify hidden patterns from unlabelled input data
- Only requires input data
- Not trained on right output
- Classify unlabelled data
11
Q
Type of ML
Reinforcement learning
A
- The system is not trained
- Make use of optimisation technique.
- Uses trial and error in algorithm to determine which action gives the optimal outcome. - e.g
search engine, online games
12
Q
Regression
A
– statistical measure used to make predictions from data by finding/learning
relationships between the inputs and outputs.
