Name :宅男的女友 (美少女) 生日 :06年 6月27日 (古希 276歲)
體: 94/97 法:12054/12054 攻擊力:0 敏捷 :0 知識 :58272
快樂 :55865 滿意 :8284 疲勞 :44 氣質 :2731 體重 :1.04
病氣 :0 乾淨 :0 食物 :54 大補丸:0 藥品 :0
◤ ◥
︵
◤︽
◥◤ ▌
▌◤▼◥▌
▌ ☽ ▌
▌ ｜ ▌
▌ ◤◥ ▌
▌ IIII█
▌ IIII█
IIII
IIII
很乾淨..飽嘟嘟..很快樂..很滿足..
Optimization is everywhere.
We will present applications of optimization to a wide range
of fields including operations management, finance, marketing, engineering,
and strategic planning, as well as operations of a university and
personal decision-making. We will also present different models and
conceptual frameworks for optimization including linear programming,
integer programming, non-linear programming, dynamic programming,
and heuristics.
Algorithms. As is traditional at MIT,
we learn the inner workings of algorithms.
It is not sufficient to say that Excel contains an algorithm that solves
linear programs. We need to know how the algorithm works.
Learning algorithms has several important implications.
First of all, some problems are easily solved, and others are intrinsically
intractable. Learning the algorithms helps to distinguish one from the other.
Secondly, understanding the behavior of an algorithm can be an important
first step in interpreting the output of the algorithm and applying it
to gain insight about the optimization problem.
Thirdly, it is only by understanding the inner working of algorithms
that we are in a position to design our own algorithms, or modify those
that exist. The goal of the models is insight, not numbers.
We build models not as a mirror of reality, but only as a partial reflection
of reality. The nature of modeling in Management Science and
Operations Research is that we approximate reality in order to
provide support for decision-making.
One useful way that models support decision-making is that they
permit managers to explore a range of scenarios in order to help in
determining which decisions are robust under a number of assumptions. In a
similar manner, one can analyze models to determine which numbers
are the most important, and which numbers can be changed with little impact
on the decision. A major theoretical tool for aiding insight is
sensitivity analysis and its variations. One caveat is the impact of
E-business on modeling. In many E-business applications,
one needs to solve thousands of models over a short period of time.
In this case, there is not time for human assessment of model outputs,
and we need to design models that are robust and trustworthy.
Performance guarantees. One of the hallmarks of optimization
(and mathematical programming) is that it provides both an optimal solution,
and also provides a succinct certificate (guarantee) of optimality.
Even when a problem is intrinsically difficult,
optimization-based techniques may provide some guarantees.
A particularly useful guarantee is a maximum distance from optimality.
Two major theoretical tools for developing bounds on the distance from
optimality are "linear programming duality" and "branch and bound."
Search techniques and heuristics. It is often the case that problems are
too intractable to be solved optimally or even nearly optimally.
In such cases, one needs to develop strategies to develop a good solution.
Such techniques are often referred to as heuristics.
We will discuss a variety of heuristics approaches including neighborhood
search, simulated annealing, tabu search, and genetic algorithms.
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 59.116.5.170