Neural network
- 2. CONTENTS: • Human Neural Network • Artificial Neural Network(ANN) • Applications of ANN • Neural Networks in software today. • Perceptron. • Backpropagation Algorithm
- 3. WHAT IS NEURAL NETWORK??? • An interconnected web of neurons transmitting elaborate patterns of electrical signals • The human brain can be described as a biological neural network 1
- 4. ARTIFICIAL NEURAL NETWORK (ANN) • A computational model based on the structure and functions of biological neural networks • A neural network is a “connectionist” computational system In 1943, Warren S. McCulloch, a neuroscientist, and Walter Pitts, a logician, developed the first conceptual model of an artificial neural network 11 2
- 5. Contd… • A true neural network does not follow a linear path. • One of the key elements of a neural network is its ability to learn • A neural network is not just a complex system, but a complex adaptive system 11 3
- 6. APPLICATION OF ANN • To perform “easy-for-a-human, difficult-for-a-machine” task. • Applications range from optical character recognition to facial recognition 11 4
- 7. PRESENT USES • Pattern Recognition • Time Series Prediction 11 5
- 8. • Signal Processing • Control 11 6
- 9. • Soft Sensors • Anomaly Detection 1 7
- 10. THE PERCEPTRON • Invented in 1957 by Frank Rosenblatt at the Cornell Aeronautical Laboratory. • The simplest neural network possible OR a computational model of a single neuron • Consists of one or more inputs, a processor, and a single output. 81 8
- 11. MULTI-LEVEL PERCEPTRON. • A computational model of a neurons • Complex to train as there are lots of processors. • Outputs generated in same manner as perceptron. • Pass through additional layers of neurons before reaching the output 1 9
- 12. THE BACKPROPAGATION ALGORITHM • Originally introduced in the 1970s, appreciated in 1976 paper by David Rumelhart, Geoffrey Hinton, and Ronald Williams. • Provides us way of computing the derivative ∂C/∂w of the cost function C with respect to any weight w. • The learning algorithm specifies the modification of weight. 1 10
- 13. WORKING • The input is passed to the input neuron. • Then the value is passed to a sigmoid function: 𝑌 𝑠𝑖𝑔𝑚𝑜𝑖𝑑 = 1 1+𝑒−cx • The result received at output end is along with error value. (Y-𝑌∗ ), where 𝑌∗ is error. 1 11
- 14. Contd… • Error is then detected and corrected using function: 1 9 12
- 15. CONCLUSION: Neural networks are suitable for predicting time series mainly because of learning only from examples. Neural networks are able to generalize and are resistant to noise. On the other hand, it is generally not possible to determine exactly what a neural network learned and it is also hard to estimate possible prediction error. 1 9 13
Editor's Notes
- #5: 3 They described the concept of a neuron, a single cell living in a network of cells that receives inputs, processes those inputs, and generates an output
- #6: 3 -> it can change its internal structure based on the information flowing through it
- #7: 2- -> (turning printed or handwritten scans into digital text.
- #8: 1-> recognition of patterns and regularities in data 2-> Neural networks can be used to make predictions. Will the stock rise or fall tomorrow? Will it rain or be sunny?
- #9: 1-> filter out unnecessary noise and amplify the important sounds. 2-> used to manage steering decisions of physical vehicles (or simulated ones).
- #10: 1-> process the input data from many individual sensors and evaluate them as a whole. 2-> detect unusual activities.
- #11: output of the network is generated by multiplying inputs by the weights are summed and fed forward through the network.
- #13: ∂C/∂w∂C/∂wof the cost function CC with respect to any weight ww
- #15: Was hi* too low or high??? Should gj be low or high??