1. 2101INT – Principles of Intelligent Systems Lecture 10

2. Neural Networks Since we are interested in creating artificial intelligence in systems, it is reasonable that we would attempt to mimic the human brain The concept of an artificial neuron has been around since at least 1943 This type of field is usually described as artificial neural networks (ANNs), connectionism, parallel distributed processing or neural computation This field is also of interested to cognitive psychologists who seek to better understand the human brain

3. Neurons A neuron is a cell in the brain that collects, processes and disseminates electric signals On their own, neurons are not particularly complex Much of the brain’s information-processing capacity is thought to stem from the number of and inter- relationships between the neurons. As such is an emergent property of the neurons, since each of its own does not have the power of the whole The human brain contains about 10 10 neurons, each on average connected to about 10,000 others

4. Neurons Signals in a brain are noisy “spike trains” of electrical energy

5. Neurons

6. The axon endings almost touch the dendrites or cell body of the next neuron – termed a synapse Electrical signals are transferred with the aid of neurotransmitters – chemicals which are released from one neuron and which bind to another Signal transmission depends on: – quantity of neurotransmitter – number and arrangement of receptors – neurotransmitter re-absorption – etc.

7. Carbon vs Silicon Elements: 10 14 synapses vs 10 8 transistors Size: 10 -6 m vs 10 -6 m Energy: 30W vs 30W (CPU) Speed: 100Hz vs 10 9 Hz Architecture: Parallel/Distributed vs Serial/Centralised Fault Tolerant: Yes vs a Little Learns: Yes vs Maybe Intelligent: Usually vs Not Yet

8. Mathematical Model of a Neuron McCulloch and Pitts (1943) 737

9. Mathematical Model of a Neuron ANNs are composed of many units A link from unit j to unit i propagates the activation of unit j (a j ) to unit i Each link also has weight W j,i which determines the strength and sign of the link Each unit computes the weighted sum of its inputs, in i : And then applies an activation function to derive the output a i : A bias weight is also present

10. Activation Functions The activation function should create a “high” output (say 1) when the correct inputs are given and a “low” output (say 0) otherwise The function also needs to be non-linear (not of the form y = mx + c ), otherwise the network as a whole would be a simple linear function, which isn’t particularly powerful Two choices are the threshold (unit step) function and the sigmoid function = 1/(1+e -x ) 738

11. Neurons as Logic Gates A single neuron can implement the three most basic Boolean logic functions (and also NAND and NOR) A single unit cannot represent XOR 738

12. Usefulness of ANNs So neurons and ANNs can be designed to exhibit particular behaviours More often, they are used to learn to recognise/classify particular patterns. Rosenblatt’s work (1958) explicitly considered this problem when a teacher is providing advice to the ANN From this we derive the term supervised classification or supervised learning It was he who introduced perceptrons: ANNs that change their link weights when they make incorrect decisions/classifications

13. Networks of Neurons Two main categories: feed-forward networks and recurrent networks Feed-forward networks are acyclic: all links feed forward in the network. A feed forward network is simply a function of its current input. It has no internal state. Recurrent networks are cyclic: links can feed back into themselves. Thus, the activation levels of the network form a dynamic system, and can exhibit either stable, oscillatory or even chaotic behaviour. A recurrent network’s response will depend on its initial state, which depends on prior inputs 738

14. A Single Layer Perceptron A network with all inputs connected directly to the outputs is called a single-layer neural network or a perceptron network 740

15. What can single layer networks do? Already seen that they can implement simple Boolean logic functions Can also represent other more complex functions, like the majority function: returns T iff more than half of its inputs are T Decision tree representation would require O(2 n ) nodes So why can’t we represent XOR?

16. Linear Seperability The output of a threshold perceptron can be described as: or as vectors: Wx > 0 This equation defines a hyperplane in the input space 740

17. Linear Seperability cont. Functions that can be divided by such a hyperplane are termed linear seperable In general, threshold perceptrons can represent only linearly seperable functions There are times when such networks are sufficiently appropriate however 741

18. How does the brain learn? Brains learn by altering the strength of connections between neurons They learn online, often without the benefit of explicit training examples Hebb’s Postulate: "When an axon of cell A... excites[s] cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells so that A's efficiency as one of the cells firing B is increased."

19. Learning What is learning? Rosenblatt (1958) provided a learning scheme with the property that: “ if the patterns of the training set can be seperated by some choice of weights and threshold, then the scheme will eventually yield a satisfactory setting of the weights” 1. Pick a “representative set of patterns” – training set 2. Expose network to this set to adjust synaptic weights using a learning rule – such as minimise the error 741

20. Learning Minimise the error, in this case sum of square error SSE: Use gradient descent, to minimise the error wrt each particular link weight. Use chain rule: 741

21. Learning cont. Using the chain rule: First term reduces to: Second term: 741

22. Learning cont. Combining these: Definitions of terms – g’() is the derivative of the activation function – in i is the weighted sum of inputs 741

23. Updating weights Having calculated the impact of each weight on the overall error, can now adjust each W j accordingly: Note that the minus has been dropped from the previous equation: +ve error requires increased output  is called the learning rate The network is shown each training example, and the weights are updated Exposure to a complete set of training examples is termed an epoch The process is repeated until convergence occurs 742

24. Learning Examples 743

25. Learning Examples 748 Graph shows total error on a training set of 100 examples

26. Multilayer Feed Forward Networks Will now consider networks where the inputs do not connect directly to the outputs Introduce some intervening units between input and output which are termed hidden units Why? 744

27. Effect of Hidden Units Call these networks multilayer perceptron networks Hidden layers remove the restriction to linearly seperable functions Using a sigmoid activation function, two hidden units can classify a ridge, 4 a bump, >4 more bumps, etc 744

28. Weight learning in MLPs Similar procedure as for a single layer, except that the error must be propagated back through the hidden layers Gives rise to back propagation learning The equations for back propagation learning are derived on pages 745-747 of the text 745pp

29. Performance comparison MLP vs Single

30. Learning in ANNs

31. Readings for Week 10 1. Russell and Norvig, Chapter 1 2. http://www.willamette.edu/~gorr/classes/cs449/brain.h tml 25pp