Module Three Mini-Homeworks (all together). Total Point Value is 0.3. 1) We have presented the database images as though each pictures size is 50x50. Thus, our experts are also 50x50. Well, that is too big for you to write down an example of an expert pattern, so for now, we will assume our sizes are 10x10. Go ahead and make up (and show it as a table of numbers) a 10x10 pattern of numbers that would qualify as a pattern for an expert (remember, you can pick from any of the Four families of patterns). 2) If you assume there are 500 positive examples (face images) and 500 negative examples (pictures of everything else, e.g., cell phones, chairs, cars, trees, walls, houses, fingers, etc.), what will the initial weight assigned to each example be? 3) Why is picking the first expert as the one with the least error, the same as picking the one with the fewest wrong? 4) If, just before the selection of the first expert, if the weights of the examples are all 0.0001, supposing some expert (call it H9) has 2500 examples wrong, what will the total for H9s error be? 5) Supposing H9 (from the previous question) is selected as the winning expert in the first selection stage, what will its Beta value be? What about its Alpha value?

6) What is the first adjustment to the weights of the 2500 examples that H9 (from the previous question) gets wrong? What is the first adjustment to the weights of the examples that H9 gets correct, what is the new adjusted value of each such weight? 7) The previous question represents the final sub-step in the first selection stage; what is going to happen next (what is the next sub-step)? 8) When an expert is selected, what is the purpose of its Alpha value? 9) When picking the second expert, give a scenario where an expert that is going to be selected (as the second) could be one that has a lower count of correct examples than others who will not be selected. Be as detailed as you need to be to create your scenario.6) What is the first adjustment to the weights of the 2500 examples that H9 (from the previous question) gets wrong? What is the first adjustment to the weights of the examples that H9 gets correct, what is the new adjusted value of each such weight? 7) The previous question represents the final sub-step in the first selection stage; what is going to happen next (what is the next sub-step)? 8) When an expert is selected, what is the purpose of its Alpha value? 9) When picking the second expert, give a scenario where an expert that is going to be selected (as the second) could be one that has a lower count of correct examples than others who will not be selected. Be as detailed as you need to be to create your scenario.