fjkdlsjflkds

7 ms·

1.
by fjkdlsjflkds 2y ago
> Non-Bayesian NN training does indeed use regularizers that are chosen subjectively —- but they are then tested in validation, and the best-performing regularizer is chosen. Thus the choice is empirical, not subjective. I'd argue t
2.
by fjkdlsjflkds 2y ago
Each cell of the 2D grid (or "neuron") is connected to (and updated as a function of) its immediate neighbours, which makes it a network/graph.
3.
by fjkdlsjflkds 2y ago
> For one, Bayesian inference and UQ fundamentally depends on the choice of the prior, but this is rarely discussed in the Bayesian NN literature and practice, and is further compounded by how fundamentally hard to interpret and choose t
4.
by fjkdlsjflkds 2y ago
Ah... gotcha. Thanks for the clarification.
5.
by fjkdlsjflkds 2y ago
It shouldn't be too much extra state. I assume that 2 bits should be enough to cover castling rights (one for each player), whatever is necessary to store the last 3 moves should cover legal en passant captures and threefold repetition
6.
by fjkdlsjflkds 2y ago
> It's not so much that as there is the net-negative side effect of having people redo the same work over and over again at their day job or whatever. It horrifies me when I think about the number of hours of people's lives was
7.
by fjkdlsjflkds 2y ago
> R's magrittr These days, base R already includes a native pipe operator (and it is literally `|>`, rather than magrittr's `%>%`).
8.
by fjkdlsjflkds 2y ago
> "an effect this large, or larger, should happen by chance 1 time out of 20" More like "an effect this large, or larger, should happen by chance 1 time out of 20 in the hypothetical universe where we already know that th
9.
by fjkdlsjflkds 2y ago
> You know that in your research field p < 0.01 has importance. A p-value does not measure "importance" (or relevance), and its meaning is not dependent on the research field or domain knowledge: it mostly just depends on ef
10.
by fjkdlsjflkds 2y ago
> It seems to me like KANs should be able to find expressions like these given experimental data. Perhaps, but this is not something unique to KANs: any symbolic regression method can (at least in theory) find such simple expressions. He
11.
by fjkdlsjflkds 2y ago
The lack of semantics associated to DC (and near-DC) components in audio data is important, and a big difference compared to image data, no doubt. I'm not sure this changes if you look at a cepstral representation (as suggested in the
12.
by fjkdlsjflkds 2y ago
> I'm not sure what do you mean by that. I mean that "trying to argue based on (your perception of) the person you are talking to, rather than what is being discussed, is a bad argumentation strategy". > How is this spe
13.
by fjkdlsjflkds 2y ago
As I mentioned, I agree with censorship when it is legitimate (ethically or morally justified), and I agree with the need for rule-of-law. It is not me that is arguing that censorship is ok when it is legal (and not ok otherwise), but Twi
14.
by fjkdlsjflkds 2y ago
The world is not black and white... there are shades of grey. Sometimes censorship is lawful and/or justified, sometimes it is not. I don't know if, in this case, it is justified or not, but it seems to be lawful (the same way t
15.
by fjkdlsjflkds 2y ago
> This feels like a distraction not an argument - it’s not relevant what happened in other countries. It's relevant, because it shows that Twitter/Musk has no problem engaging in state-mandated censorship, as long as it the man
16.
by fjkdlsjflkds 2y ago
> If he did that , it would be just as deplorable [...] No need for hypotheticals. He did do that (this is an easily-verifiable fact [1][2][3]). > [...] the censorship they were claiming to oppose. The thing is that this is clear
17.
by fjkdlsjflkds 2y ago
> These are in no way equivalent. e.g. the first amendment only protects you from the government not from private organizations (if anything them deciding to publish or not to publish your content is an expression of freedom of speech an
18.
by fjkdlsjflkds 2y ago
Exactly. It is perfectly legal for a private entity (such as Twitter) to engage in censorship, as they regularly do so. So, the argument that "we can't do that, because that would be illegal" doesn't really fly. Furtherm
19.
by fjkdlsjflkds 2y ago
> You can if there is a venue for that. If the government is behaving in arbitrary and authoritarian way trusting it to do the right thing is a bit silly... I assume that the judge in question used a specific criminal or civil law to j
20.
by fjkdlsjflkds 2y ago
If you cannot appeal (and you probably can't, since this was a judicial order by the Supreme Court), then you have to comply (or face the consequences of ignoring judicial orders). If the argument is that it is illegal to "censor&
21.
by fjkdlsjflkds 2y ago
> This is unconstitutional in Brazil per article 5 of the 1988 constitution, so X refused the orders. This is unconstitutional according to their interpretation of the (very extensive and vague) article 5 of the 1998 constitution, mayb
22.
by fjkdlsjflkds 2y ago
This. Unfortunately, I couldn't read beyond the first page, since it keeps crashing my (desktop) browser. Would be nice to have a button to stop/remove all animations in the page, so I could actually read the rest.
23.
by fjkdlsjflkds 2y ago
This works nicely: https://en.wikipedia.org/wiki/Harmonic_table_note_layout Nice to play chords with a single (fat) finger, and you can go through the circle of fifths by moving vertically.
24.
by fjkdlsjflkds 2y ago
Thanks for taking the time to clarify my confusion. It's not that there is "little variance left to explain", but actually that (no matter what) there will always be too much variance left to be explained, when the response
25.
by fjkdlsjflkds 2y ago
(After re-reading the blog post with more care...) you are right, and thanks for the correction. Either way, the point stands... the improvement in using a full linear model (that predicts 0.45 or 0.55, depending on state) is marginal compa
26.
by fjkdlsjflkds 2y ago
An affine model, yes (as I mentioned in my comment). A linear model, no ;) But, anyway, seems like I interpreted things incorrectly.
27.
by fjkdlsjflkds 2y ago
This, of course, will depend on how you code your variables, but if you try to fit a null, intercept-only and predictor-only model, you get this as residual variance: > data <- data.frame(state = c(0, 1), pref = c(0.45, 0.55)) > su
28.
by fjkdlsjflkds 2y ago
> whether an effect is statistically significant ("does the effect exist?") and whether it's significant ("does the effect matter?") The more accurate terms to describe this is whether an effect is significant
29.
by fjkdlsjflkds 2y ago
> The R^2 used with a linear model requires a constant term, in this case the constant term or bias explains a lot about preferences (almost 50/50) so there is less information available for the slope term. This explains the paradox
30.
by fjkdlsjflkds 2y ago
OP does not (seemingly) claim that tinygrad can't compute hessians, only that a first-order optimization method was the only thing tried. Also, as a quasi-newton method, L-BFGS does not require explicit (pre-)computation of the hessian

More ›