matjet

6 tháng trướcdiscuss

At this time, the competition is soon finishing - with no models having succeeded. Given the incentives for top labs, and the short time needed for a successful automated solution, we can make a reliable upper bound on the capability of current models - better than any normal benchmaxed datasets.

What I would like to see is an easier version of this same format.

matjet

6 tháng trướcdiscuss

Look what they need to mimic a fraction of [the power of having the logit probabilities exposed so you can actually see where the model is uncertain]

matjet

3 năm trướcdiscuss

From "Clickbait is bad, and I don't do it." to "I am working to efficiently clickbait people." Reminds me of:

https://www.youtube.com/watch?v=S2xHZPH5Sng

matjet

3 năm trướcdiscuss

I think you misunderstand induced demand. The increased demand when you add more capacity, simply means even more utilisation is achieved -> increased productivity. This is not a waste of time.

The correct comparison should be comparing for each area and situation, which type of transport investment results in the greatest utility. The objective is not to minimise traffic, but to maximise peoples ability to get where they need to be.

matjet

3 năm trướcdiscuss

Lore dump: This is not a coincidence, as the characters themselves refer to the chidori being named due to sounding like thousands of birds. The alternate name used by kakashi, raikiri, is related to https://en.wikipedia.org/wiki/Tachibana_D%C5%8Dsetsu

matjet

3 năm trướcdiscuss

Is this a sarcastic joke, or is it a serious comment? I honestly cannot tell.

matjet

3 năm trướcdiscuss

The latter option is closest, but neither is quite right. It would have ~known~ that the problem asked, combined with a phrase for a 15 line limit has associations with a length of 12 lines (perhaps most strongly 12, but depending on temp it could have given other answers). From there it is constrained to (complete) solutions that lead to 12 lines, from the several (partial) solutions that already exist in the weights.

matjet

3 năm trướcdiscuss

Sharing of the model weights should now be fine. There is a 12 hour old post for the original pull request that was dismissed on the basis of being unlikely to be considered; that links the preferred url.

EDIT; I may have misunderstood the reviewers acceptance. The post should probably be deleted.

matjet

3 năm trướcdiscuss

Gaussian blur is essentially acting as a low pass filter. An IR filter does not strictly destroy information in the filtered spectrum components, but does attenuate their power.

Given a perfect blurred image, reconstruction is possible - however due to the attenuation, these high frequency components are ~sensitive~.

Apart from quantisation effects [you mentioned which limits perfect de-convolution], adding a little AW Gaussian noise(such as taking a photo of the image from across the room) after the kernel is applied obliterates high frequency features.

Recovery when noise is low (plus known glyphs) is why you should not use Gaussian blur followed by print screen to redact documents. Inability to recover when there are artifacts and noise is [part of] why cameras cannot just set a fixed focus [at whatever distance] and deconvolve with the aperture [estimated width at each pixel] to deblur everything that was out of focus.

TLDR for readers, It is unlikely to recover sufficient detail via de-convolution here.

matjet

3 năm trướcdiscuss

Hear Hear! This deserves to be reiterated given the statements of information recoverability made in this thread.

matjet

3 năm trướcdiscuss

To someone only weakly aware in this area, can you explain what is wrong with the statement? From a cursory view, these protocols indeed seem to normally carry the mentioned payloads.

matjet

4 năm trướcdiscuss

While 1991 does seem too recent for the text, 1961 is not possible given it makes several references to later years, up to 1965/1968.

I am curious if this article has been scanned by a machine that attempts to use OCR to improve quality - at the expense of occasional mistakes. "dress" in place of "chess" looks close in font appearance, but far to mistype on a keyboard.

matjet

4 năm trướcdiscuss

Can someone please elaborate on the relationship between archive links and spam? Why is this particular auto-flagging done?

matjet

4 năm trướcdiscuss

Patent trolls that are hated look like: You develop package to do X from first principles, then get sued because someone patented using a known algorithm for the purpose of X.

Copyright working in a supported/non hated way: You develop a package to do X by cribbing off someone else's package X. They sue you for stealing their work, not to make money off you. Situation at hand is case 2, hence the lack of interest in financial gain.

Why is this case 2, when it does not always reproduce the copyrighted works exactly? Situation: You realise that rather than cribbing off of one persons package X, you can crib off two other package X's and mix/average their contents. Scale this to 100's of packages.

Eventually, ML should avoid this by developing to work from first principles, writing in it's own style, with public code used only for validation of it's ability to understand and write code.

matjet

4 năm trướcdiscuss

Texas has the same pattern as Europe; higher percentages of solar and wind, resulting in an overdependence on gas.

matjet

4 năm trướcdiscuss

The correct analogy here would be implementing spam filtering by blocking large segments of email addresses. Eg, dropping mail from all non microsoft/gmail domains (as a nuisance reduction measure!), with predictable impact on smaller providers and self hosted email.

matjet

4 năm trướcdiscuss

To add a datapoint:

In this video https://www.youtube.com/watch?v=T8q3zrCYMRw, they managed to differentiate most high end violins (4/5 tests - 10 Violins).

Incidentally, the only mistake was with one of the two Stradivari examples.

matjet

4 năm trướcdiscuss

I do not see the use for this, over the standard access methods.

    people = {
        "Diane": 70,
        "Bob": 78,
        "Emma": 84
    }

    people.get("Bob")
    # 78

The common usecase is:

    for key in people.keys():
        people.get(key)

    keys = people.keys()
    # dict_keys(['Diane', 'Bob', 'Emma'])

    for key in keys:
        keys.mapping[key]

What is the advantage?

matjet

5 năm trướcdiscuss

Numerical simulations that I know of, all use steps or iterations of some type. Even if we start with theoretical infinite/perfect precision of the double pendulum, the numerical simulation will diverge from (ideal frictionless!) reality.

Two simulations of chaotic systems (starting identically) with different step sizes will always diverge (The difference in eventual positions does not stabilise as steps are made smaller). For this reason, I am not even sure if infinitesimal steps would avoid divergence from (ideal) reality. Plus y'know, the whole issue of a simulation with infinitesimal steps never making ANY progress, regardless of how fast it runs.

Therefore, I conclude that infinite degrees of precision is not the issue or solution for numerical explanation of chaotic behavior.

matjet

5 năm trướcdiscuss

Multidimensional array construction is something I had looked forward to in Julia. I am not convinced that the approach taken in julia1.7 compares favorably with other language implementations (Ignoring R) In my view the numpy syntax has more clarity for this task. [[[1,2],[3,4]],[[5,6],[7,8]]] compared with [1 2;3 4;;;5 6;7 8]

It is not immediately clear why [1,2,3,4] is equivalent to [1;2;3;4], but [1,2,,3,4] (and [1,2;3,4] vs [1 2;3 4] ect) is not equivalent to [1;2;;3;4]. For creating a 3d slice, I expected that ";", ";;", ";;;" would each refer to incrementing a specific dimension. Eg, It seems intuitive that if you can create a 2d matrix with [1 2;3 4], then you should be able to make a 3d tensor with [1 2;3 4;;5 6;7 8]

Submissions

LLaMA models licence change to Apache 2.0 approved

comments