M12 - Unsupervised Learning

Question 1

What are the categories of machine learning?

Answer 1

• Supervised learning
• Unsupervised learning
• Reinforcement learning

Question 2

Explain Supervised Learning and give examples.

Answer 2

• labeled data
• feedback
  EX. Regression, SVM, Neural Networks
  Object recognition

Question 3

Explain Unsupervised learning and give examples.

Answer 3

• no labeled data
• no feedback

Ex. K-mean, self-organizing maps

Clustering

Question 4

Explain Reinforcement learning and examples.

Answer 4

• reward based learning
• increase cumulative discounted rewards

Ex. Q-learning, SARSA, TD-learning, DQNs

Robot Navigation

Question 5

What is the motivation behind using unsupervised learning?

Answer 5

• no hand-labeling data required but we can still perform classification
• find structure (or clusters) in the dataset based on the idea: similar inputs should be located close to each other
• do not need backprop to adjust the weight

Question 6

What kind of models can we create with unsupervised learning?

Answer 6

• find structures (clusters) in the data
• detect anomalies
• create auto-association
• reduce the dimensionality

Question 7

When should we use unsupervised learning?

Answer 7

• when we have a lot of sensory data and we can not hand-label them
• we do not have a rewards function to guide the behavior of robot
• we can not structure a score (loss) function to train the network

Question 8

How does k-means work?

Answer 8

• Partitions data into k number of mutually exclusive clusters
• how well a point fits into a cluster is determined by the distance from that point to the cluster’s centre

Question 9

What are the convergence criteria for k-means?

Answer 9

1. Number of specific iterations
2. Until the centres moves with a threshold value

Question 10

How do we find correct numbers of K?

Answer 10

• Visualize data
• Domain knowledge
  Run for different number of K then select the best one

Question 11

What is competitive learning?

Answer 11

• paradigm different that supervised learning and reinforcement learning
• the aim is not to map input and output by minimizing a cost function
• no explicit reward function to tune the behaviour of network
• form a unit (or neuron) which can be activated for similar inputs

Question 12

How does competitive learning work?

Answer 12

• Similar inputs will activate the units that are close to each other
• the activation of the neuron will increase if the distance between the neuron weights and input vector is closer

Question 13

What is the are the Pseudo-code steps of competitive learning?

Answer 13

1. Initialize the weights for each unit
2. Derive Euclidean distance for each unit for an input
3. Select the unit which has the smallest Euclidean distance as the winner unit
4. Update the weights of the winner unit

Question 14

What happens if the learning rate = 0?

Answer 14

No updates will be made to the model’s parameters during the training process

Question 15

What if the learning rate = 1?

Answer 15

Model’s parameters will be updated by a large step at each iteration

Question 16

What does competitive learning allow you to do?

Answer 16

• cluster the input
• reduce the dimensions
• extract features in a unsupervised way

Question 17

What is SOM?

Answer 17

Self organizing map = is a variation of competitive learning

Question 18

What does SOM do?

Answer 18

• the units close to each other (by weights) respond to similar inputs and have similar weights

Question 19

What are the limitations of SOM?

Answer 19

• the initialization of the weights may lead to significant differences in the final formation of the SOM
• training duration increases with dimensions of the network
• different SOMs can find different similarities for the same inputs

Question 20

Competitive learning vs. Self-organizing maps

Answer 20

Competitive Learning:
- Identifies features/prototypes through neuron competition.
- Focuses on individual feature identification.
- Used for feature selection and identification.

Self-Organizing Maps (SOMs):
- Creates a low-dimensional representation of data while preserving topology.
- Emphasizes cluster formation and visualization.
- Used for clustering and visualization tasks.

Question 21

What are the use cases of SOM’s?

Answer 21

• Epigenetic Robot Architecture (ERA)
• Self-Organized Internal Models Architecture (SOIMA)

Question 22

Explain Epigenetic Robot architecture.

Answer 22

• the units of the architecture are self-organizing maps
• each SOM can be used to process different modality input
• missing modality representation can be constructed

Question 23

Explain Self-Organized Internal Models Archtectur.

Answer 23

• the unit of the architecture are self-organizing maps
• implementation of inverse and forward models in the same architecture
• missing modality representation can be constructed
• sensory-motor mapping