![AAA
HARIKA
38
To determine how to predict exam performance based on the number of
hours of sleep we get and the number of hours we study the previous day
• Collect a lot of data, and for each data point
x = [x1 x2]T
• Record the number of hours of sleep (x1), the number of hours spent studying (x2)
and whether performed above or below the class average
• Our goal, then, might be to learn a model h(x, θ) with parameter vector
• θ = [θ0 θ1 θ2 ]T such that](https://image.slidesharecdn.com/introduction-220823052231-174feeca/85/Introduction-pptx-38-320.jpg)
![HARIKA
39
• Guess that the blueprint for our model
h(x, θ) is as described as a linear
classifier that divides the Cartesian
coordinate plane into two halves.
• Then, need to learn a parameter vector
θ such that our model makes the right
predictions (−1 if performance below
average, and 1 otherwise) given an
input example x
• This model is called a linear perceptron
• Then it turns out that by selecting
θ =[ −24 3 4 ]T
• The learning model makes the correct
prediction on every data point](https://image.slidesharecdn.com/introduction-220823052231-174feeca/85/Introduction-pptx-39-320.jpg)
Introduction.pptx
- 1. Amrita School of Engineering, Bangalore 19ECE354 - Deep Learning Ms. HARIKA PUDUGOSULA Lecturer Department of Electronics & Communication Engineering Artificial Neural Networks
- 2. HARIKA 2 fig: Image from MNIST handwritten digit dataset A zero that’s algorithmically difficult to distinguish from a six
- 3. HARIKA 3 Text book defination of Deep Learning - • Deep learning is a subset of a more general field of AI called machine learning, which is predicated on this idea of learning from example. • In machine learning, instead of teaching a computer a massive list of rules to solve the problem, we give it a model with which it can evaluate examples, and a small set of instructions to modify the model when it makes a mistake.
- 4. 4 HARIKA
- 5. 5 HARIKA
- 6. 6 HARIKA
- 7. 7 HARIKA
- 8. 8 HARIKA
- 10. HARIKA 10
- 11. 11 HARIKA
- 12. 12 HARIKA
- 13. 13 HARIKA
- 14. 14 HARIKA
- 15. 15 HARIKA
- 16. 16 HARIKA
- 17. 17 HARIKA
- 18. 18 HARIKA
- 19. 19 HARIKA
- 20. 20 HARIKA
- 21. 21 HARIKA
- 22. 22 HARIKA
- 23. 23 HARIKA
- 24. HARIKA 24
- 25. 25 HARIKA
- 26. 26 HARIKA
- 27. 27 HARIKA
- 28. 28 HARIKA
- 29. 29 HARIKA
- 30. 30 HARIKA
- 31. 31 HARIKA
- 32. • Deep learning-driven wireless communication for edge-cloud computing: opportunities and challenges https://journalofcloudcomputing.springeropen.com/articles/10.1186/s 13677-020-00168-9 • Deep Reinforcement Learning for Multiagent Systems: A Review of Challenges, Solutions, and Applications https://ieeexplore.ieee.org/document/9043893 HARIKA 32
- 33. 33 HARIKA
- 34. HARIKA 34 Expressing Linear Perceptons as Neurons
- 35. HARIKA 35
- 36. HARIKA 36
- 37. HARIKA 37 • A bias acts exactly as a weight on a connection from a unit whose activation function is always 1 • Increasing the bias increases the net input to unit • Fixed threshold can also be used for the activation function instead of bias weights
- 38. AAA HARIKA 38 To determine how to predict exam performance based on the number of hours of sleep we get and the number of hours we study the previous day • Collect a lot of data, and for each data point x = [x1 x2]T • Record the number of hours of sleep (x1), the number of hours spent studying (x2) and whether performed above or below the class average • Our goal, then, might be to learn a model h(x, θ) with parameter vector • θ = [θ0 θ1 θ2 ]T such that
- 39. HARIKA 39 • Guess that the blueprint for our model h(x, θ) is as described as a linear classifier that divides the Cartesian coordinate plane into two halves. • Then, need to learn a parameter vector θ such that our model makes the right predictions (−1 if performance below average, and 1 otherwise) given an input example x • This model is called a linear perceptron • Then it turns out that by selecting θ =[ −24 3 4 ]T • The learning model makes the correct prediction on every data point
- 40. HARIKA 40 • An optimal parameter vector θ positions the classifier so that we make as many correct predictions as possible. • Most of the time these alternatives are so close to one another that the difference is negligible. • If not the case , what needs to be done? • How do we even come up with an optimal value for the parameter vector θ in the first place? • Solving this problem requires a technique commonly known as optimization • An optimizer aims to maximize the performance of a learning model by iteratively tweaking its parameters until the error is minimized. • What happens when they is an uneven distribution of positions ?
- 41. HARIKA 41
- 42. HARIKA 42 **singular neurons are strictly more expressive than linear perceptrons**
- 43. HARIKA 43 Every linear perceptron can be expressed as a single neuron, but single neurons can also express models that cannot be expressed by any linear perceptron