234 Matching Annotations
  1. Last 7 days
    1. Bellman backup operator

      Given a valuation and a state, returns the expected return provided we follow the action that maximizes the next reward plus the valuated value of the next state.

    2. return

      Return at time \(t\)

      The smaller the discount factor, the smaller is the relative weight of the far future rewards.

    3. Monte Carlo and -soft Policies

      It may make sense to decrease epsilon gradually while training to explore less and exploit more (refine the policy for the successful runs).

    4. Monte Carlo with Exploring Starts

      Can we make the policy stochastic and choose each action with a probability that scales with its expected return?

    5. greedy action

      Under what conditions do we get stuck in a local optimum?

  2. www.ciirc.cvut.cz www.ciirc.cvut.cz
    1. Český ústav robotiky a kybernetiky

    1. regularization constantC

      Finite C implies forces w to be in a hypersphere of finite radius R that depends on C.

    2. (w∗,b∗) = argminw∈Rn,b∈R(12‖w‖2︸︷︷︸penaltyterm+Cm∑i=1max{0,1−yi(〈w,φ(xi)〉+b)}
    1. Vapnik-Chervonenkis dimension

      \(D_{VC}(H) = max_n . \exists x \in X^n . \forall y \subseteq x . \exists h \in H . h(x) = y\)

    2. `0/1(y,y′) = [[y6=y′]

      We assume 0-1 misclassification loss.

    3. Rn

      Space of feature vectors

    4. ULLN for finite hypothesis space

      and bounded loss image (bounded by [a, b])

    1. average return of 490

      There are 500 steps in each episode.

      Is gamma equal to 1.0 in evaluation?

    2. Assignments: Best solution counts. There is no penalty for incorrect solution submission. The students can ask for increasing the submission count limit (50 by default).

    3. monte_carlo

      Train either in ReCodEx or offline.

      If I can't get it running in an hour, I can write to MS and ask for advice.

    4. states

      Generalized states: R^4:

      1. Cart position
      2. Cart velocity
      3. Pole angle
      4. Pole angular velocity

      For this task the states are discretized by binning (8 bins in each dimension; 8^4 states in total).

    5. --steps steps of policy evaluation/policy improvement

      We improve the policy steps times.

    6. --iterations applications of the Bellman equation

      After iterations evaluations, we improve the policy once.

  3. Oct 2019
    1. sup

      "Uniform" because we care about all strategies in H.

    2. limm→∞P(R(hm)−R(hH)≥ε)= 0


      \(\forall \delta > 0, \epsilon > 0. \exists m_0 \in N. \forall m \geq m_0. P(R(h_m) - R(h_H) \geq \epsilon) < \delta\)

      In general:

      \(lim_{m \rightarrow \infty} f(m) = 0 \Leftrightarrow \forall \delta > 0. \exists m_0 \in N. \forall m \geq m_0. f(m) < \delta\)

    3. estimation error

      How to reduce estimation error?

      • Increase number of samples m.
      • Improve the learning algorithm to utilize T^m more efficiently.
      • Shrink H to exclude strategies better than h_m.

      Enlarging H typically leads to worse generalization, so estimation error increases.

    4. R(hm)

      R(h_m) is a random variable, because h_m is a random variable, because T^m is a random variable.

    5. YX

      All functions \(h: X \rightarrow Y\)

    6. approximation error

      How can we reduce approximation error?

      Enlarge H so that it includes better rules.

    1. Average value per assignment: 11 points

    2. I can ask for a justified deadline extension before the deadline passes.

    1. Q=n+1Q+nα(R−nQ)

      Exponential moving average

      \((1 - \alpha) Q_n + \alpha R_n\)

    2. Incremental Implementation

      New estimate for this machine is old estimate plus relative current reward divided by number of times we tried this machine.

      Cf. gradient descent

    3. uniformlyrandom action with probability .ε1−εε

      \epsilon: Rate of exploration

    4. discrete


    5. won 10 out of 11 StarCraft II games against two professional players

      Unfair interface advantage

    6. 9h

      Very high computational demands

    7. ches

      Can beat the previously best program after 3h of playing

    8. beat 60 professional

      Nickname: master

      Online game server, the program did not reveal it was a bot

    9. AlphaGo

      Why is go interesting?

      • High branching factor
      • Difficult to write a good state evaluation heuristic
    10. to rule them all

      Plays all the games

    11. human-normalized mean: 623.0%

      Questionable information value because some games can be exploited a lot

    12. 29 games out of 49 comparable or better to professional game players

      A separate agent trained for each game

    13. first


    14. Trial and error learning

      Learning in animals

    1. VGGNet

      conv3-64: receptive field 3x3, stride 1, 64 output channels

      To resolve boundary: pad with zeros

      MP: max pooling layer. Window 2x2, stride 2, aggregate with max.

      Deep layers: more channels, smaller resolution, large transitive receptive field, more abstract features

    2. negative log-likelihood of class probabilities (a.k.a. cross entropy)

      Surrogate loss function (or objective function) for 0-1 loss

      Why? So that it is differentiable.

    3. Loss function

      If we used 0-1 loss, the empirical risk would be the error rate.

    4. class probabilities

      To choose the class, we take the class with the highest predicted probability.

    1. Empirical Risk Minimization

      Optimization problem

    2. Testing: confidence intervals

      Filip: Is the estimation based on Hoeffding inequality commonly used in practice (research)?

      A: Yes.

    3. Hoeffding inequality

      Strength: Valid irrespective of the underlying distribution

      What distribution maximizes the error?

    4. 1−2e−2lε2(b−a)2=γ

      This equality binds the variables \epsilon, \gamma, l and (b-a).

    5. [a,b]

      The variables are bounded interval by an interval \([a, b]\).

    6. Example (rolling a die):μ= 3.5,zi∈[1,6],ε= 0.5.
      • Left graph
        • Red line: True expectation (3.5)
        • Blue band: \epsilon band around assumed expectation
      • Right graph:
        • Red curve: Portion of experiments that estimate a mean outside the blue band in the left graph
        • Blue curve: Upper bound of the empirical established by Hoeffding inequality
    7. how the interval widthεdepends onlandδ

      3 interdependent variables:

      • \(\epsilon\)
      • \(l\)
      • \(\delta\)

      3 possible tasks:

      1. Given l and \delta, minimize \epsilon.
      2. Given \epsilon and l, maximize \delta.
      3. Given \epsilon and \delta, minimize l.
    8. (RSl(h)−ε,RSl(h)+ε)

      Confidence interval

  4. Sep 2019
    1. NP-complete


      NP complete or NP hard

    2. Seminars

      Theoretical assignments every other week published one week in advance.

      No evaluation, just discussion.

    3. Written exam

      90 minutes. Assignments similar to those discussed in the labs.

    4. The course will run irrespective of the number of students. The lecture format will only be held if at least approx. 5 students come consistently.

    1. maximum likelihood estimator


      You should know this already.

    2. unsupervised learning

      Note that the validation set must still be labeled.

    3. Bayes inference rule

      Note that this only works for classification.

    4. inf

      Is this infimum?

    5. how strong can they deviate from each other?

      Example: For 0-1 loss, using Chebyshev inequality, using the loss bound of 1, we get the bound: $$\frac{1}{\epsilon^2 m}$$

    6. drawn fromp(x,y)

      The evaluation distribution must match the application distribution. In other words, we evaluate the inference with respect to p.

    7. i.i.d.

      independent identically distributed

    8. R(h)

      In case of 0-1 loss in classification, this is the error rate of h.

    9. E(x,y)∼p


    10. minimises the loss

      The distribution of (x, y) will be specified later.

    11. related by an unknownjoint p.d.f.p(x,y)

      We assume that x and y are interdependent.

    12. loss function`:Y×Y →R+

      Simplest loss for classification: 0-1 loss: l(a,b) = 1 iff a != b

    13. real valued variable

      possibly a vector

    1. Labs/Seminars: Thursday 9:15-10:45, 11:00-12:30 and 12:45-14:15, all in KN:E-112

      Capacity: 30 people

    2. Boris:

      Don't hesitate to contact us.

      Don't hesitate to ask questions during the lectures.

      Immediately after each lecture we will go to the cafeteria where you can contact us privately.

    1. practical labs

      Each task: Around 10 points plus up to 3 bonus points

    2. before the class


      a week ahead

    3. Seminar

      Introductory lab session. Neither theoretical, nor practical.

  5. Aug 2019
    1. public key

      To have a personal certificate generated by CIIRC, follow the instructions in the wiki.

    2. Everyone with CIIRC account can access master node of the cluster via ssh at cluster.ciirc.cvut.cz

      Credentials: CTU username and CIIRC password

  6. Jul 2019
    1. All other usage is reserved.

      Does this mean that modification of Vampire for research purposes is prohibited?

    1. max

      According to the paper (Cohen 1998), maximizing the permutation score is an NP-hard problem. The paper proposes a 2-approximation greedy algorithm.

  7. Jun 2019
  8. May 2019
    1. Moc emocí, málo informací a argumentů.

    2. Vyber si niečo, za čím sa budeš hnať, až dokým ťa nezradí vlastné telo.

      Myslím, že tento postoj je zároveň součástí ideologie autora tohoto textu.

    3. Ľudí, ktorí chápu že spôsob, akým funguje naša spoločnosť, je chorý! Ľudí, ktorí vidia riešenia, nie problémy! Ľudí, ktorí kladú zaujímavé otázky a majú zaujímavé odpovede! Ľudí, ktorí sa neboja povedať svoj názor v spoločnosti, ktorá nové názory odsudzuje, no zároveň je na nich závislá!!!

      Apel na emoce

    4. si môžeme písať čo chceme

      Tohle mi připomíná manipulativní heslo Parlamentních listů:

      Nikdo nám nediktuje, o čem smíme psát.

      Kritická analýza: Jeden svět na školách

    5. Nikto nás nevlastní.

      Zatím se mi zdá, že qTube vlastní níže podepsaný autor tohoto textu.

    6. Epidémia depresie, ktorú zažívajú “rozvinuté štáty” je spôsobená nezdravým spôsobom myslenia a inherentne nesprávnymi konceptami, ktoré považujeme za pravdivé.

      Chybí zdroje.

    7. Ak nás za našu kontroverziu a excentrické vystupovanie odsúdite, je to v poriadku. Je to spôsob, ktorým filtrujeme tých správnych ľudí.

      Co když vás odsoudím za manipulativní rétoriku?

    1. HP Color LaserJet 6040 MFP

      Color LaserJet cm6040 MFP

      Viz screenshot v bodu 6 níže.

    2. System -> Správa -> Tisk

      Ubuntu 18: Settings -> Devices -> Printers -> Additional Printer Settings...

      Zdroj: Ask Ubuntu

    3. Server -> Nová -> Tiskárna

      Ubuntu 18: Add

  9. Apr 2019
    1. Of course, it’s okay for them to fly – Emma Thompson jetted first-class from LA to London to lecture us plebs about all our eco-destructive holidaymaking.

      This reminds me of the numerous shots of Al Gore at an airport or on an airplane in the 2006 environmentalist film An Inconvenient Truth.

    1. a property of a person that is true for Andrew but not for Betty

      Such property is implicit: = @ andrew

    2. thf(heated_coffee_mix,axiom, ! [S: syrup] : ( ( heated_mix @ coffee @ S ) = coffee ) ).

      A heated mixture of coffee and any syrup is still coffee.

    3. thf(hot_mixture,axiom, ! [B: beverage,S: syrup] : ( hot @ ( heated_mix @ B @ S ) ) ).

      A heated mixture is hot!

    4. thf(heated_mix_type,type,heated_mix: beverage > syrup > beverage).

      A heated mix is the application of heat to the mixture of a beverage with a syrup.

    5. thf(heated_mix,axiom, heated_mix = ( ^ [B: beverage,S :syrup] : ( heat @ ( mix @ B @ S ))) ).

      A heated mix is the application of heat to the mixture of a beverage with a syrup.

    6. thf(mix_type,type,mix: beverage > syrup > beverage).

      A mixture of a beverage with a syrup is a beverage.

    7. thf(heat_type,type,heat: beverage > beverage ).

      A heated beverage is a beverage.

    8. thf(coffee_type,type,coffee: beverage).

      Coffee is a beverage.

  10. Mar 2019
    1. same density


      The density in the current test cases is no more than 0.2.

    2. You can call the function solveDA(rule, da_params) multiple times

      According to songzy12, it is necessary to call the function at least once to get a score. See discussion in the forum.

    3. The edge weights are all 1.

      There may be multiple edges between any pair of nodes.


      I think you should accept multiple edges between the same pair of nodes.

    4. T = (actual execution time of your script) - (DA API wait time) + (only the annealing time within DA)


      "Total execution time(ms)" in the email is the time we are using for scoring that counts towards 180 sec. limit.

      -- Meaning "Total execution time(ms)" in the email is the T, and "DA API wait time" is already excluded from there.

  11. Feb 2019
    1. Each column-j cannot have any duplicate number

      Moreover, this constraint ensures that each column has at least one number of each in \(K\). The constraint ensures that each column contains each number in \(K\) exactly once.

  12. Jan 2019
    1. Část A: Vytvoření bootovacího USB

      Instrukce v této sekci jsou platné pouze pro Windows. Pokud máte Linux, následujte instrukce v jednom z následujících návodů:

    1. it is most likely that these effects come from the breathing techniques

      Why do you think so?

    2. An example of why testing should always be blinded!

      How exactly would that help in this case?

  13. Dec 2018
    1. 16. Jak vysoká bude podpora v nezaměstnanosti a jak dlouho ji mohu pobírat?

      maximální výše PvN činí 0,58násobek průměrné mzdy v národním hospodářství za první až třetí čtvrtletí kalendářního roku předcházející kalendářnímu roku, ve kterém byla podána žádost o PvN.

      Zdroj: Základní poučení

      Průměrná mzda v národním hospodářství v roce 2017: 28 761 Kč

      Zdroj: Sdělení MPSV

      Maximum v roce 2018: 0,58 * 28 761 = 16 681,38 Kč

  14. Aug 2018
  15. Jul 2018
  16. May 2018
    1. You can not only apply it to classes, but also to software components and microservices.

      The principle can also be applied to source code files.

  17. Apr 2018
  18. Feb 2018
    1. you’ll need to know up front that you want your commit applied into multiple places, so that you can place it on its own branch

      More specifically, you'll need to know up front all the places you want your patch commit applied into so that you can determine where to start the patch branch from.

    2. if you can anticipate where a commit may/will be need to applied

      This is an important assumption.

      The way I understand it, cherry picking is intended to be used in case of unanticipated migration of code.

    3. its own branch

      The patch branch should start at a common ancestor of all the target branches. Since the broken code is in all the target branches, it must be in at least one of their common ancestors.

      If we don't start from a common ancestor and merge the patch into all the target branches, at least one of the target branches will get some extra change along with the patch.

  19. Jan 2018
    1. call to a function defined in a DLL file

      Is it really possible to get an "undefined variable" compilation error from a function defined in a DLL?

      First of all, the calls to the DLL are only resolved by the linker. Second, I don't think the linker analyzes the definitions of the functions at all. Third, I don't think there are any "variables" in native binary code; there are just memory addresses and registers in play.

  20. Dec 2017
    1. The PowerShell equivalent to the Unix which command is called get-command.

      Also try the command where in Windows command prompt. It does not only run in PowerShell.

  21. Nov 2017
    1. Individuals and interactions over processes and tools

      Do not give up personal interaction in favor of a tool-supported process. "Když tomu nerozumím, tak se jdu zeptat."

    1. /dev/cdrom on /mnt/cdrom type iso9660 (ro,nosuid,nodev)

      VMware Workstation Player 12:

      /dev/sr0 on /media/ubuntu/VMware Tools type iso9660 (ro,nosuid,nodev,relatime,uid=1000,gid=1000,iocharset=utf8,mode=0400,dmode=0500,uhelper=udisks2)

    1. Go to Control Panel > Uninstall a Program > Turn Windows features on or off to turn off Hyper-V.

      This step was sufficient on my computer.

    1. exists as long as the feature is in development

      When the development of a feature takes a long time, it may be useful to continuously merge from develop into the feature branch. This has the following advantages:

      • We can use the new features introduced in develop in the feature branch.
      • We simplify the integration merge of the feature branch into develop that will happen at a later point.
    1. ignorance rather than malice

      When the author prefers to explain a phenomenon by ignorance rather than malice, he is applying Hanlon's razor.

  22. Jul 2017
    1. .

      This character should be escaped by a backslash. The complete command would then be:

      strings $PWD/bin/myapp | egrep '\.gcda$'
  23. Jun 2017
    1. We must ensure that precompiled headers are disabled when the code is instrumented.


  24. May 2017
    1. Every shared library has a special name called the ``soname''. The soname has the prefix ``lib'', the name of the library, the phrase ``.so'', followed by a period and a version number that is incremented whenever the interface changes (as a special exception, the lowest-level C libraries don't start with ``lib''). A fully-qualified soname includes as a prefix the directory it's in; on a working system a fully-qualified soname is simply a symbolic link to the shared library's ``real name''.Every shared library also has a ``real name'', which is the filename containing the actual library code. The real name adds to the soname a period, a minor number, another period, and the release number. The last period and release number are optional. The minor number and release number support configuration control by letting you know exactly what version(s) of the library are installed. Note that these numbers might not be the same as the numbers used to describe the library in documentation, although that does make things easier.
  25. Mar 2017
  26. Feb 2017
    1. Ale je mi jasné, že to nejde, že bych jim to tam nakonec asi narušovala.

      Jestli bys to narušovala záleží na povaze akce. Tvá účast v mužském kruhu by, myslím, vyžadovala, aby účastníci potlačili svou tendenci považovat Tě za ženu. Mohl by to umožnit rituál (samotný mužský kruh je rituálem; třeba by to stačilo) v kombinaci s vědomým odhodláním účastníků.

  27. Nov 2016
  28. Oct 2016
    1. Very likely your Linux partitions are Ext3.

      Nowadays ext4 is used in Ubuntu.

    1. Jiří Ovčáček to obratně zdůvodnil špatným počasím při přistání v Bratislavě.

      Proč bych měl pochybovat o pravdivosti tohoto zdůvodnění?

  29. Sep 2016
    1. it’s easy to argue by induction

      It may be easy to argue by induction, but what is actually going on?

      Let's take a group of three children for a larger example: Abe, Ben and Cindy. Let's denote the situation "Abe has a clear forehead and Ben and Cindy have muddy foreheads" as [OXX], and so on.

      Before the teacher tells them anything, Abe knows that either [XXX] or [OXX] (because he sees the mud on Ben's and Cindy's foreheads). If [OXX] were the case, Ben would know that either [OXX] or [OOX]. If [OOX] were the case, Cindy would know that either [OOX] or [OOO].

      This picture demonstrates the situation.

      The teacher's announcement makes it common knowledge that [OOO] is not possible. Then if Abe supposed [OXX] and provided that Ben supposed [OOX], Cindy would be sure that [OOX] (meaning she has mud on her forehead). After the first call, she doesn't stand up so we know that if Abe supposed [OXX], in that idea Ben could no longer suppose [OOX] and would be instead sure that [OXX] (meaning he has mud on his forehead).

      This example suggests that the children don't need the general notion of common knowledge to efficiently reason whether they have muddy foreheads. It suffices to use any chain "Abe considers possible that Ben considers possible that Cindy considers possible etc.". that visits every child exactly once in an arbitrary order.

    2. the emperor is naked

      This is a reference to the short tale The Emperor's New Clothes.

    3. “would you like to come up for some intercourse?”

      Incidentally, this straightforward approach is taken by Phoebe towards Chandler in the episode 5x14 titled "The One Where Everyone Finds Out" of the TV show Friends:

      I'm really looking forward to you and me having sexual intercourse.

      As pointed out by James Miller in his video introducing the concept common knowledge, this episode contains a nice informal demonstration of the concept.

    4. such things
      • preserving the romantic fantasy
      • a face-saving way to back out later
    1. téměř pětina z 1200 oslovených rodičů obou pohlaví udala, že kdyby se měli dnes rozhodovat znovu, děti by neměli

      „Wenn ich mich heute noch einmal entscheiden könnte, würde ich keine Kinder mehr bekommen wollen.“ [19 % žen, 20 % mužů]

      Source: Regretting Parenthood

    2. téměř všichni (až na tři procenta) tvrdili, že své děti milují

      „Ich liebe mein Kind / meine Kinder.“ [96 % žen, 93 % mužů]

      Source: Regretting Parenthood

    1. Until Let’s Encrypt fixes their bullshit, the CAcert certificate stays.

      As of now (2016-09-20), xaymar.com actually does use a certificate issued by Let's Encrypt. It would be very interesting to read a follow-up article about the reasons that lead to this switch. Sadly, I haven't been able to find one.

  30. Aug 2016
    1. Visible identification marks


      If the applicant has a visible identification mark (such as birthmark or tattoo) please type in the information here. Otherwise, type “NONE”.

    2. Guides for filling in this form:

    1. Find approximate values for the first three iterates,x1,x2, andx3, using the seedx0= 10.

      $$x_0 = 10, x_1 = 27, x_2 = 59.13, x_3 = 72.499293$$

    2. f(x)

      $$f(x) = -\frac{3}{100} \, {\left(x - 50\right)}^{2} + 75$$

    3. What isf(40)?


    4. What isf(20)?


  31. Jul 2016
    1. he may move the one checker a total of eight spaces to an open point, but only if the intermediate point (either three or five spaces from the starting point) is also open

      What happens if the intermediate point is occupied by a single opponent's stone?

  32. Jun 2016
    1. fixing issues before you submit is infinitely preferable to having your solution challenged or having it fail during system testing

      Why is that? As far as I understand, when I submit an incorrect solution, I am only risking contributing a 50 point bonus to successful challengers of the solution.

    1. g++ -Wall -W -O2 -s -pipe


      -s Remove all symbol table and relocation information from the executable.

  33. May 2016
  34. Feb 2016
    1. 490,-

      Platí se nejspíš až za druhou a každou další kartu. Jedna karta je k MůjÚčet a Konto G2.2 zdarma.

  35. Jan 2016
    1. eyebrows


      Songs written with one idea in mind have been known to mutate into something completely different if I hear an 'optional vocal inflection' during rehearsal. I'll hear a 'hint' of something (often a mistake) and pursue it to its most absurd extreme.

      The 'technical expression' we use in the band to describe this process is: "PUTTING THE EYEBROWS ON IT." This usually refers to vocal parts, although you can put the eyebrows on just about anything.

    1. require 'active_record/connection_adapters/postgis_adapter/railtie'

      The following worked better in my setup:

      require 'active_record/connection_adapters/postgis/railtie'
    1. with another instance as its receiver

      An instance can call private methods of another instance of the same class.

    1. Double or Nothing

      If you get this after you achieve your daily goal, today will not count towards the Double or Nothing streak.

    1. In desperation people reach for the sledgehammer; if their kids won't listen to them, maybe they'll listen to God.
    1. addictive but unrewarding pastimes

      Playing video games, watching porn, reading blog posts etc.

    2. life actually is short

      if you have kids.

      Can you convince someone who does not have kids that life is short?

    3. things that matter

      for the future; in the long-term view. This corresponds to optimizing overall happiness using exploration rather than exploitation. There also seems to be certain intuition tied to the concept of "mattering" though that is no easy to explain as "maximizing overall happiness".