TB10 B&B Lecture 3; How are neurons connected to serve useful functions? 0.0 / 5 ? PsychologyTB10 B&BUniversityNone Created by: mint75Created on: 23-05-16 14:48 What is the core premise of neural computation? A specific pattern of activity in one pattern of neurons induces a specific pattern of activation in another 1 of 18 What are the three features of Selfridges (1959) pandemonium model of letter recognition? Feature 'demon' -> cognitive 'demon' -> decision 'demon 2 of 18 What is the concept of generality? Neural networks can be connected to transform ANY set of input patterns into ANY set of output patterns 3 of 18 What is a connections weight? A concept analogous to the total strength of all inhibitory and all excitatory connections between the neurons 4 of 18 What is meant by monotonic? As there is more input there is more output 5 of 18 Which is correct? (sum of)Activation of input neuron x weight of input 6 of 18 What is typically used in modelling? 3 layer architecture 7 of 18 What is meant by learning? Finding a set of connection weights to make the network useful by making systematic changes to the weights in response to input patterns 8 of 18 What are the two criteria for generality? Each input pattern maps onto only ONE output pattern and there are enough units (unlimited) and layers of units (4) 9 of 18 Do neural networks start with initially random weights? Yes 10 of 18 In supervised learning, what happens after each input pattern? Connections are weakened/strengthened as to reduce contributions to discrepancy between observed and desired output 11 of 18 What variant of learning does deep learning use? Supervised learning 12 of 18 Can deep learning approach and exceed human knowledge on tasks such as face and object recognition? Yes 13 of 18 Which output neuron wins in competitive learning? Whose input weights best match the pattern of activity in the input units 14 of 18 What ensures that each output unit has a fair chance of responding to each input pattern? The total weight of connections to each output unit is limited. If others strengthened others weakened 15 of 18 What would happen if there were no competition rules? The same units would win again and again and become stronger but would not be selective for diff inputs. Other units would never win 16 of 18 In a feature space graph, what do arrows represent? Weights going into an output unit 17 of 18 In a feature space graph, what do clusters represent? Input activation patterns 18 of 18
Comments
No comments have yet been made