5. Do you want fries with that? Well, if you\'re regularly going to a small rest

Question

5. Do you want fries with that? Well, if you're regularly going to a small restaurant, they recognize you and know what you've been ordering in past visits-so they know exactly what additional goodies to suggest to entice you to spend more money. Suppose you have been hired to improve computer software that uses face recognition in a fast food restaurant to match customers with their past purchases and prompt the person taking the order to suggest the most profitable additional goodies the database says that particular customer is likely to buy. Unfortunately, the first version of the software is too slow - the customer has usually concluded their purchase before the system can make its first suggestion. You need to get a 2X speedup for the system overall to be viable. The code spends 60% of that time executing face recognition code, 30% in database lookup checking sales records, and 10% in an expert systen inferring what should be suggested to maximize expected profit. You think you can speed-up the face recognition quite a lot, but no other aspects of the system software or hardware can be changed. Is it theoretically possible to meet the performance goal that way or are you doomed to fail? Why? (Hint: Amdahl.) Toi 6. 100% As you're looking for your first job after graduating, you find out from a friend of a friend about an exciting new start-up company. You interview there and discover they have found a way to create a program that will do something truly wonderful and, if it works, this program will change the world for the better and make everybody in the company very wealthy. They figure it will take about 3 years to develop the program and bring it to market. The catch is that their little mock-up used to test the ideas suggests that current computer hardware is about 5X too slow to make the program work as intended. Should you seriously consider joining this company? Why or why not? ical

Explanation / Answer

5. Facerecognition is capturing the face image , extracting the compact feature and matching against face templates stored in the database. This facerecognition modul is  parallelizable and parallelization will speed-up the system. theoraticaly it is possible to speed up by n time doing n parallel matching.

6. As per Moor's law the computation power of hardware doubled every year. The said program required 3 year to developed. In that three years the computation power of hardware increased to 2X to 4X to 8X. So whene the programm is ready to run we have the required powerfull hardware tu run the program. So I will seriously consider to join the company.

Related Questions

Navigate

• Browse (All)
• Browse 5
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.