document.write('
    \x0a \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a
  1. \x0a Processing\x0a \x0a \x0a
    \x0a

    “Processing is an open source programming language and environment for people who want to program images, animation, and interactions.”

    \x0a
    \x0a \x0a
    2. \x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a
  2. \x0a Model Selection Homepage\x0a \x0a \x0a
    3. \x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a
  3. \x0a \x0a
    Ph.D. Candidate!
    \x0a \x0a \x0a
    \x0a

    I am a Ph.D. candidate since 16:30 today! Hurray!

    \x0a
    \x0a
    4. \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a\x0a \x0a\x0a \x0a
  4. \x0a
    \x0a “Focusing on important questions puts us in the awkward position of being ignorant. One of the beautiful things about science is that it allows us to bumble along, getting it wrong time after time, and feel perfectly fine as long as we learn something each time.”\x0a
    \x0a \x0a \x0a
    \x0a The importance of stupidity in scientific research by Martin A. Schwartz.\x0a
    \x0a \x0a
    5. \x0a \x0a\x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a
  5. \x0a How to Publish in Top Journals\x0a \x0a \x0a
    \x0a

    by Kwan Choi.

    \x0a
    \x0a \x0a
    6. \x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a
  6. \x0a Unlabeled data: Now it helps, now it doesn’t\x0a \x0a \x0a
    \x0a

    by Aarti Singh, Robert Nowak, and Xiaojin Zhu.

    \x0a
    \x0a \x0a
    7. \x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a \x0a \x0a
  7. \x0a
    \x0a \x0a
    \x0a \x0a \x0a
    \x0a

    Why Achievement is not normal?

    \x0a
    \x0a \x0a
    8. \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a
  8. \x0a \x0a
    Relations between Probability Distributions
    \x0a \x0a \x0a
    \x0a

    (credit: I found these links here.)

    \x0a
    \x0a
    9. \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a
  9. \x0a \x0a
    Sobolev Embedding Theorems, Besov Spaces, etc.
    \x0a \x0a \x0a
    \x0a

    First we need the definition of Hölder and Sobolev spaces (Wikipedia).

    \x0a

    The idea of Hölder spaces is that the k-th derivative of the functions are alpha-Lipschitz. The idea of Sobolev spaces is that the k-th derivate of the functions are in Lp space.

    \x0a

    Then we have different Sobolev inequalities. These inequalities basically say which Sobolev space can be embedded in another one. Of course, if two spaces are measured with the same Lp norm, the one with higher degree of smoothness can be embedded in the less smooth one. But if we measure one with Lp and measure the other with Lq (p!=q), then the relation is more complicated and depends on p, q, smoothness degrees and the dimension n of the space R^n.

    \x0a

    Then we have interesting Gagliardo-Nirenberg-Sobolev inequality which says for diff. function u: R^n —>R, ||u||_Lp* <= C || D u ||_Lp where p* is the Sobolev conjugate of p, and D is the differentiation operator.

    \x0a

    Now, we have Besov spaces. As far as I understand, there are different ways to define it: Fourier (or Wavelets)-based approach and moduli of smoothness approach.

    \x0a

    For example, see DeVore, Sharpley, “Besov Spaces on Domains in R^d” (Section 2).

    \x0a

    Jan Schneider, Function Spaces with Varying Smoothness, Ph.D. Dissertation.

    \x0a

    It starts by Fourier-analytical approach to define Besov spaces in R^n (Definition 2.1), reviews some embedding theorem (Theorem 2.2), and equivalent of some norms (Theorem 2.3) in Besov spaces. Then it defines the same thing for domains \\Sigma \\subset R^n, and proves similar results. Afterwards, it generalizes the same result to the case where the smoothness measurement changes. The motivation is that this way is better to capture pointwise singularities. Again, similar properties hold.

    \x0a

    And finally a few other references:

    \x0a

    Frayssee and Jaffard, “The Sobolev Embeddings are Usually Sharp”.

    \x0a

    I am not sure if understand the paper correctly, but this is my take from its Introduction: For a Sobolev space W^s(Lp)(R^n), there exists a function belong to that space which has a Holder exponent of s - d/n (the Holder exponent more or less show the maximum pointwise smoothness of the function). Now, the question is whether this function belonging to the aforementioned Sobolev space is an exception or actually most functions in that Sobolev space has this Holder exponent of s - d/n.

    \x0a
    \x0a
    10. \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a \x0a
  10. \x0a \x0a
    Generalized Representer Theorem
    \x0a \x0a \x0a
    \x0a

    Scholkopf, Herbrich, Smola, “Generalized Representer Theorem”.

    \x0a\x0a

    The theorem is quite general: for an arbitrary loss function L(.) that depends on data x_1,…,x_m, and regularizer in the form of g(||f||) with g(.) as strictly monotonic increasing, we have the usual expansion in the form of f(.) = \\sum_{i=1}^m a_i k(.,x_i) .

    \x0a

    Also the Remark 6, biased regularization, is quite interesting. I thought people had not considered it before!

    \x0a
    \x0a
    11. \x0a \x0a\x0a \x0a\x0a \x0a\x0a \x0a \x0a \x0a \x0a \x0a\x0a \x0a \x0a
');