Artificial Neural Networks
Tips
Er worden ook bijvragen gesteld, zorg dat je dus iets meer weet dan het antwoord op de vraag zelf.
Vragen
We hebben 50 voorbeeldvragen gekregen waaruit we er 2 gesteld gaan krijgen.
(over deze pagina: Indien je het antwoord op een vraag weet of gewoon pagina referenties vul ze dan in aub. Onenigheid over antwoorden wordt best in de discussion page beschreven. Indien je niet zeker genoeg bent van je antwoord (waar geen reden toe is, iets halfjuist is beter dan niets) kan je het ook gewoon in de discussion page zetten.)
Lecture 1
Explain differences between digital computers, neural networks and the human brain.
Explain the difference between learning, memorization and generalization.
Explain the operating principle of the neuron and the perceptron and the adaptation of the weights
In which areas is it useful to apply neural networks and where is it not useful?
Lecture 2
Which kind of activation functions can one use for neural networks? What are advantages and disadvantages with respect to learning?
Explain the difference between feedforward and recurrent nets. Give some examples.
What is the difference between supervised and unsupervised learning? Give examples.
Explain the discrete and continuous perceptron and delta learning rule. What is the difference and why?
Explain the difference between linearly separable and linearly nonseparable tasks. Which tasks can be handled by a single layer neural network or by a multilayer neural network ?
Explain the backpropagation algorithm and the choice of the parameters and early stopping rule.
Lecture 3
What is an attractor neural network and how does it work? Give an example.
Does the Hebb learning rule guarantee pattern stability in the Hopfield model? Explain.
Is the loading capacity of the Hopfield model bounded? Explain.
Explain the concept of energy function in neural networks?
Explain the projection method.
Can one describe a temporal sequence of patterns with the Hopfield model? Explain.
Lecture 4
How can one solve optimization problems using the Hopfield model? Give an example.
Explain the energy function of the travelling salesman problem.
Design a network for the weighted matching problem.
What is reinforcement learning? Give some examples.
Explain the recurrent back-propagation algorithm.
Explain the associative reward-penalty algorithm.
Explain learning with a critic. Can it be applied for the control of a plant?
Lecture 5
What sort of general tasks can one perform with unsupervised learning?
Explain the standard competitive learning rule.
Explain vector quantization and learning vector quantization. What is the difference?
Explain the adaptive resonance theory algorithm.
What is a self-organizing map? Give an example.
Are hybrid learning schemes useful? Explain with an example.
Lecture 6
Explain Oja's rule in the framework of unsupervised Hebbian learning.
Give the connection between Sanger's rule and principal component analysis.
What is the aim of pruning and construction algorithms? Give an example.
Explain the tiling algorithm.
Lecture 7
What is the relation between the Bayesian and maximum likelihood approach?
Explain: Bayesian learning treats the issue of model complexity differently than cross-validation does.
Why does one favour small values in Bayesian learning of network weights? Explain.
What are the basic ideas implemented by support vector machines? Explain.
What are the motivations for fuzzifying the perceptron rule? How can one do this?
Lecture 8
Explain the problem of echo cancellation and the use of adaline networks to overcome it.
What is a cellular neural network and how does it work ?
What are typical application areas of cellular neural networks?
What is a CNN template?
Lecture 9
What are strong points and drawbacks for applications of neural networks to time-series prediction?
Discuss time-series prediction competitions and results obtained by neural networks.
Discuss the use of neural networks in the alvinn vehicle control system.
Explain the use of backpropagation in control applications and what is backpropagation through the plant ?
Explain the difference between feedforward and feedback neural control.
Lecture 10
Discuss different neural control strategies for controlling an inverted pendulum.
Discuss the design of a neural network for controlling a truck backer-upper.
Discuss neural networks for use in ovarian cancer classification and prediction.
Discuss the use of neural networks in fraud detection applications.
Explain the receiver operating curve, its use and its role in comparing classification systems.
Bijvragen
(Algemene bijvragen kan je hierzetten, indien ze een uitbreiding op een andere vraag is zet je ze best daarbij.)