EIE4512 Mid-term Review

Week 1 bits and intensity

  • spatial resolution
  • interpolation

Week 2 Contrast Stretching

  • gamma (choose a filter for image)
  • histogram (pseudo code and api)
  • contrast
  • dynamic range (num of distinct intensity of an image)
  • cumulated histogram (from dynamic range) equalization
  • given different cumulated histogram, choose the curve of equalization

Week 3 Histogram Matching

  • how to match a histogram
  • linear spatial filter (correlation and convolution)
  • under what condition correlation and convolution are the same?
  • smoothing (lowpass: gaussian, average, medium…) and sharping (highpass, high boost, )

Week 4 Fourier Transformation / IFT

  • impulse train
  • sampling theorem (the frequency of sampling should be two times of …)
  • DFT / IDFT (given one function to calculate the DFT/IDFT)
  • no matter IDFT and IFT will result in complex, spectrum, log, magnitude/phase angle
  • translation insensitive, rotation sensitive
  • filters (ideal lowpass, Butterworth, gaussian, corresponding highpass filters)

Week 5 Image Restoration

  • degradation model, the formula of additive noise
  • noise distribution
  • mean filter (contra-harmonic, salt and pepper)
  • for example given a salt noise, how to define a contra-harmonic filter
  • alpha-trim filter, when will become mean filter
  • adaptive filter (local, three situation and different situation)
  • given example of local variance and global, which filter should you choose
  • adaptive medium filter
  • bilateral

Week 6 Edge Detection

  • smooth (why should we do smoothing)
  • trade off between localization and detection
  • Hague transformation
  • line detection
  • interest point (which is a good interest point)
  • why we are so interest in the corner detection
  • harris corner detection (familiar with the whole process:loss function/ structure tensor identify what it is)

Week 7 Brightness Consistency Equation

  • brightness consistency equation
  • Lucas – Kanade
  • interactive refinement
    • coarse-to-fine optical flow

关于埃航坠机事件

3月10日Ethiopian Airlines Flight 302于起飞不久后发生坠机,近日引发轩然大波,各种相关的新闻、报道、调查层出不穷。作为Air Crash Investigation 的(伪)真实粉丝,我觉得还是可以有一点评论的。

我觉得媒体不应当在官方调查结果出来之前用一种看起来极为确定的方式来试图解释事故原因。事实上飞机在起飞后失控的可能状况有很多,波音公司在研发改良款飞机时也一定进行过诸多调查。在历史上媒体乱带风向,官方报告反而没人信的情况也不是一次两次。倒不是说媒体是动不动就要搞一个大新闻,即便是出于调查真相的好心,媒体之力也难以得出足够客观的结果——毕竟空难调查需要专业的团队、专业的设备、充足的时间、细致的考察才能得到准确的结论。在当下这个时候人云亦云并无好处。

此外我觉得我的同学参与写作的坠机之后,埃塞俄比亚的大非洲航空野心何处安放倒是取了一个不错的观察角度,谈了谈事故之后对航空公司以及整个非洲航空市场的潜在影响。我觉得媒体就应该找这些角度做一些独立的调查。

论如何取一个博客名字

最近个人的感情突然开始变得丰富,突然有了一些很有意思的想法。

事实上地球上有70亿人,生物学上能和你匹配的可能顶多有个20亿,而我这一辈子能遇到的,我觉得10000个差不多了。

How many bits to distinguish these 10000 individuals?

About 14 bits!

所以事实上假如谈恋爱像一个猜数字游戏的话,其实不需要几轮问题就可以得出结论的。当然你需要选择一个好问题。

至于如何取博客名字,我希望这个名字是全英文可读的(这和生成一个好记的密码条件差不多)

其次我希望这个名字是unique among 7 billions of 地球人口的。 所以我至少需要 log2(7,000,000,000) = 30bits 换算下来八个字符差不多了。我不希望以后面临像iPv4一样的窘境,但事实是,目前这个名字在google上是搜不到结果的,并且任何一个我所知道的域名后缀都没有注册这个名字。And that is good.

6 Modal and Anti-Luck Epistemology

this is a reading summary of the book “THE ROUTLEDGE COMPANION TO EPISTEMOLOGY”

It is a defense to the Gettier Problem, by adding extra requirements for JTB to be a knowledge: belief should be true “out of luck”.

The defense says, it is not the way Smith’s think that make the belief to be true.

What do you mean by good luck in Smith’s case?

Because the justification is not “connected” to the fact

Definition

Justification

  1. avails herself of sufficient relevant evidence,
  2. reasonably thorough reflective examination of evidence
  3. the evidence is appropriately connected to the fact
  4. forms a belief on the basis of evident

Luck

  1. justified in believing
  2. the justification is not appropriately connected to the fact
  3. is nevertheless true

Defense

Making sure the justification is appropriately connected! – Difficult (lottery case)

Because the belief is not connected to the fact

Definition

Duncan Pritchard

“A matter of luck that the agents’ belief is true”

Parallel universes, where the belief is not universally true.

Modal epistemology might say that a belief counts as knowledge only if it is true not only in the actual world, but also in a certain proportion of worlds.

Problem

How sufficient is a sufficient proportion?

Distinct way of giving expression to the anti-luck intuition

  • Sensitivity: whether or not the belief is true in fact is not ensured by the way Smith believes — in other words, whether the belief in his mind is sensitive to the truth
  • Safety: under some circumstance, his belief can be just false

Sensitivity Theories

Robert Nozick

test if one’s belief is sensitive to truth: considering the nearest possible world in which p is false, and consider if S still believes that p. Only sensitive belief could be knowledge

Robert Nozick (revised)

S knows that p iff:

  1. p is true
  2. S believes p via believe M
  3. if p weren’t true and S were to use M to infer that p is not true; and
  4. if p were true and S were to use M to infer that p is true.

Keith DeRose (argues)

Safety Theories

Even in the parallel universe of the same situation, Smith’s belief could be false due to some other reason.

Ernest Sosa

belief is safe iff: S would believe that p only if it were so that p

(updated)

If S knows a contingent proposition, p , then if in most nearby possible worlds using the same way to form the belief, S believes that p only when p is true

(even stronger)

if S knows a contingent proposition, p, then in nearly all (if not all) nearby possible …

Quiz

The difference between safety theory and sensitivity theory.

Safety theories and sensitivity theories are both moral epistemic theory to specify “how sufficient the proportion of the other world to have the connection of the belief/justification and the truth, that can made the belief knowledge”

But two theories work in different direction:

  • in sensitivity theories, they discussed the nearby worlds where the belief is false and analyses the justification of the protagonist in that case.
  • in safety theories, they discussed the (most of) nearby possible worlds where S still believes in p, and if the belief is still true in that case.

GEB2004 期中总结

我和世界的interaction,使得我形成了belief,那些被认为好的belief会被称之为knowledge。

为什么要这样的认可?

因为主观和客观总是有距离的。我的belief和世界的fact亦是有距离的。人作为群居动物,需要这样的共识,需要identify好的source of information。knowledge is such a mark for “good information”, or “good belief”

我们如何认可一个belief是一个knowledge?

That’s theory of knowledge

So what is knowledge?

1. Gettier Problem

一个关于什么是Knowledge的波澜

Knowledge 是一个在日常生活中广泛运用,后来才变成术语的概念。

之前大家认为一个belief是knowledge的criteria是:

  • belief
  • justified
  • true

两个Gettier提出的反例(Gettier’s cases):

  • 求职的Smith
  • 福特汽车的故事

其他关于Gettier Problem的case(Gettier’s case) 大概还有几十个吧。

2. Skepticism

也许没有知识可言?

我需要首先知道自己不是一个BIV(缸中之脑)才能知道自己有没有手

可是我不会知道我是不是BIV啊!

Modus tollens

额外:关于epistemology上的学术讨论

往往都是围绕Gettier case的讨论

3. Peer disagreement

如何考虑自己和别人认识上差异的

几个理论:

  • 取平均
  • 死活不听
  • 我要是很确定的话,就应该坚持
  • 两边都不应该继续那么相信

4. Externalism 和 Internalism

一个justification从何而来,发自内心的认同justification是不是重要?

Definition

Internalism: One should know the justification from their inside. (understand the justification in heart, perhaps a higher order of personal justification)

Externalism: That is an unnecessary requirement, which may resulted in skepticism. (may result in endless loop in seeking for justification?)

5. Reliabilism

只要你的justification看起来很reliable就够了

Reliabilism 是 Externalism 的一个case

Generality Problem: 一个reliabilism的致命问题

Reliabilism的形成过程是否有一个unique的归因?毕竟他们可能包含不同的reliability啊!

这个分类的原则必须要确定,否则reliablism就没法玩了。

6. Anti-luck epistemology

分析为什么Gettier Case不是knowledge而只是luck

knowledge不应该是due to luck的

可以讨论平行宇宙(其他nearby world)中的情况,来以此分析luck

为什么上述case is due to luck? What does “luck” imply?

  • sensitivity theory: 在别的世界中要是这个belief是false的呢?
  • safety theory: belief要是在别的世界还是true,belief是不是始终不变呢?

分歧:是不是necessary bala-bala。。。

7. Fundationlism

Fundationlism and Coherentism are theories of the structure of epistemic justification – how can a knowledge be justified.

8. Coherentism

Similar to the above topic.

9. Contextualism

The truth condition of knowledge attributions vary with the contexts in which the knowledge attribution is made.

10. Epistemic Justification

How to justify

11. Experimental Epistemology

 

12. Feminist Epistemology