Skip to main content
VictorZurkowski's user avatar
VictorZurkowski's user avatar
VictorZurkowski's user avatar
VictorZurkowski
  • Member for 10 years, 10 months
  • Last seen more than a month ago
comment
Checking elementary proofs with proof checkers
Is the day when these "proofs" are rendered in a natural language coming soon?...
comment
Primitive element theorem without building field extensions
@David Looking for a proof of the primitive element theorem lead me to your answer. You observe that the Bezout resultant is a non-zero polynomial in t, and conclude that it is not the zero polynomial function (of t), which is not always the case. Is it possible to clean the argument?
revised
Loading…
revised
Loading…
revised
Loading…
comment
Wasserstein distance in R^d from one dimensional marginals
@Neyman identified a problem with my argument.
Loading…
revised
Loading…
comment
Is $C^{\infty}(\mathbb{R}^{m+n})$ a flat module over $C^{\infty}(\mathbb{R}^{m})$?
@Pietro, Do you see how $sum_i r_i m_i$ is an element of $<r_1,\dots,r_d>_{C^\infty(X)} \otimes_{C^\infty(X)} C^\infty(X\times Y)$? I am working in "coordinates", where the abstract tensor product is made concrete as $<r_1,\dots,r_d>_{C^\infty(X\times Y)}$, and in these coordiantes, the zero function is the zero element. I hope that helps.
comment
Is $C^{\infty}(\mathbb{R}^{m+n})$ a flat module over $C^{\infty}(\mathbb{R}^{m})$?
@Yemon, my gut feeling is that $C^\infty({\bf R})\otimes_{\bf R} C^\infty({\bf R})$ is not $C^\infty({\bf R^2})$. I will think of a proof. As per my answer, my first step was to verify that $<r_1,\dots,r_d>_{C^\infty(X)} \otimes_{C^\infty(X)} C^\infty(X\times Y) = <r_1,\dots,r_d>_{C^\infty(X\times Y)}$ by checking that any $C^\infty(X)$ bilinear map factors, etc (which is easy), and then I realized I was extending scalars, so I dropped the original argument and appealed to extending scalars.
comment
About martingales induced by iterative processes
When $f=0$, $\{X_i\}_i$ is a random drift, and it does not converge. Usually, in iterative methods, the learning rate $\eta$ is assumed to decrease. To second @Mateusz comment, I also doubt the sequence can converge with fixed $\eta$ in general.
Loading…
Loading…
comment
Сoincidence of discrete random variables
'@Iosif I see. You are right, sir
comment
Сoincidence of discrete random variables
'@Iosif, $E(\eta) \le E(E(\xi| \eta) ) = E(\xi) \le E( E(\eta|\xi)) = E(\eta) $, therefore the inequalities are equalities. Then $E(\xi|\eta) - \eta$ is a non-negative function whose integral is 0, hence $E(\xi|\eta) = \eta$ a.e.; likewise $ E(\eta|\xi)) - \xi$ is a non-negative function whose integral is 0, so $ E(\eta|\xi)) = \xi$ a.e.
comment
Сoincidence of discrete random variables
As @Iosif mentions, the argument applies with any $g$. One can take $g(x) = x$.
awarded
comment
Checking $f(x_1,y_1)f(x_2,y_2)-f(x_1,y_2)f(x_2,y_1) \ge 0$
It may be negative. The logarithm of such an expression (or any number) is positive if and only if the expression is greater than or equal to 1.
Loading…
revised
Loading…