This quiz contains multiple-choice problems on the basics of feedback neural networks, pattern storage network analysis, stochastic networks, Boltzmann machine and the analysis of auto-associative neural networks.
What is the objective of linear auto-associative feedforward networks?
To associate a given pattern with itself
To associate a given pattern with others
To associate output with input
None of the above
Is there any error in linear auto-associative networks?
Yes
No
If the input is ‘a(l) + e‘, where ‘e’ is the noise introduced, then what is the output in the case of an auto-associative feedback network?
a(l)
a(l) + e
Could be either a(l) or a(l) + e
e
If the input is ‘a(l) + e‘, where ‘e’ is the noise introduced, what is the output if the system is accretive?
a(l)
a(l) + e
Could be either a(l) or a(l) + e
e
If the input is ‘a(l) + e‘, where ‘e’ is the noise introduced, then what is the output if the system is interpolative?
a(l)
a(l) + e
Could be either a(l) or a(l) + e
e
What property must a feedback network have for it to be useful in storing information?
Accretive behaviour
Interpolative behaviour
Both accretive and interpolative behaviour
None of the above
What is the objective of a pattern storage task in a network?
To store a given set of patterns
To recall a give set of patterns
Both to store and recall
None of the above
Is it true that linear neurons can be useful for applications such as interpolation?
Yes
No
For what purpose is energy minima used?
Pattern classification
Pattern mapping
Pattern storage
None of the above
What is the capacity of a network?
The number of inputs it can take
The number of output it can deliver
The number of patterns that can be stored
None of the above
The number of desired patterns is __ of basins of attraction.
Dependent
Independent
Dependent or independent
None of the above
What happens when the number of patterns is more than the number of basins of attraction?
False wells
Storage problem becomes hard problem
No storage or recall can take place
None of the above
What happens when the number of patterns is less than the number of basins of attraction?
False wells
Storage problem becomes hard
Neither storage nor recall can take place
None of the above
What is a Boltzmann machine?
A feedback network with hidden units
A feedback network with hidden units and probabilistic update
A feed forward network with hidden units
A feed forward network with hidden units and probabilistic update
How can false minima be reduced in case of recall error in feedback neural networks?
By providing additional units
By using probabilistic update
Can be either probabilistic update or using additional units
None of the above