dizzant

Exactly correct. Quantizing is like rounding. This is the same notion as in quantum mechanics: the values aren’t any real number (a huge input space), but rather integer multiples of some constant (a small output space). Quantization is “rounding” to the nearest allowed output.

To build on the example used in the article, you could quantize an image by dropping the lower half of the bits of every pixel value. This discards half the information by mapping the larger input image space (e.g. 8 bits per pixel) to a smaller output image space (e.g. 4 bits per pixel), effectively scaling/rounding the original values.

The fact that higher frequencies can be quantized more intensely with relatively little image quality loss is a completely separate idea from quantization itself. In this context, high frequency masking as described in the article could be accurately described as a spatially dependent quantization to zero-bit output values. But the quantization part is the mapping, not the shape of the mask, and I think the article is misleading about the distinction.

In real codecs, there usually isn’t a hard boundary like a mask, but instead an empirical table of how many bits to keep based on the perceptual value of the information at each frequency.

I’m not saying that quantization doesn’t lose information. Instead, I’m saying that every step to lose information is not quantization. Functions on the preimage (clipping, masking) are not quantization.

I'm surprised that 10 years later, the article still teaches an incorrect conceptual understanding of "quantization".

The context heavily implies that quantization is "discarding information" generally (e.g. clipping, masking). The wikipedia article he links in the footnote disagrees, and so do I:

> quantization is the process of mapping input values from a large set (often a continuous set) to output values in a (countable) smaller set, often with a finite number of elements.

Quantization is a function over values, or the image itself. Clipping and masking, by contrast, are functions over the space/frequency dimensions, i.e. the pixel indices ("preimage"). A rule like "apply a circular mask" or "drop the first quadrant" does not depend on the values!

The most extreme example of quantization is a threshold filter. Regardless of location, output "1" if the pixel is brighter than X, or "0" otherwise. This depends only on the value, not on the location. He also misses the opportunity to point out that chroma subsampling is another application of quantization, this time to chroma values (drop the low-order bits).

Codecs like H.264 and JPEG use spatially-varying quantization rules to effectively discard more information at high frequencies, so his explanation of H.264 is spot on, but a non-expert encountering quantization for the first time would probably come away with the wrong idea. For an article explicitly targeted to non-experts that goes out of its way to introduce the term, it's a pretty important oversight.

> Only if those bold reforms weren't in the interests of the people they serve.

This is my point exactly: the creative process is impossible to do perfectly. If you want to find potential policies to enact, you need to examine very many policies that don’t work and aren’t in your constituents best interest. Those get left on the table and not advanced to actual congressional actions, sparing the legislator scrutiny for their creativity.

Some creative solution-seeking will happen in a bipartisan way through negotiation. I understand the desire to see the whole process, but removing the veil would change the behavior the participants. If every word were subject to partisan scrutiny from the wings, how could congresspeople possibly get anything productive done? See as evidence congressional committee hearings, in which the members spend significant time grandstanding for C-SPAN sound bites instead of productive questioning.

> When their behavior is conducted in secrecy behind closed doors the public doesn't have the ability to hold them accountable for it

This is obviously not true. Every bill has publicly documented sponsors and goes through a lengthy, public process with several rounds of procedural votes, committee approvals, amendment debates, etc. At some point there must be a minimum granularity of attribution. Track Changes from blank page to finished law would be useless anyway, because it would show some unpaid staffer actually typed most of the changes. Even Git history uses commits, and very mature projects like the Linux kernel revise their patches rather than keeping a detailed history of the patch creation. The champion who landed it and their reviewers are credited with the patch, as also in Congress.

> Often most of that work is done entirely for them by corporations and lobbying groups

The drafting and ideation? Sure! I wouldn’t want it any other way. It’s much better that we have an entire industry of people studying the law, finding solutions, and advocating their preferred options, rather than a group of Philosopher Kings making the law in isolation. You’re also dramatically downplaying the effort of coalition building, balancing constituent pressure, and passing procedural hurdles that is required to take a bill all the way from draft to law.

I think we may agree that provenance of bill text originating from outside of Congress could potentially be more transparent. But then again, the relevant interest groups are usually very well known by the time a bill reaches the floor anyway because that is their whole purpose, and they want to take credit for their victories. And “Monsanto drafted these provisions” is pretty far from the vision of absolute transparency and zero congressional privacy that you seem to be suggesting.

Do you have other specific ideas to increase transparency? If bill sponsor attribution, public debate and amendment, and recorded voting record are insufficient to understand a legislator’s actions, what would you suggest instead?

Transparency has a cost. There are several good reasons, but my favorite is intellectual freedom. Policy makers would be much less likely to explore bold reforms or risky bipartisan compromises if their draft quality notes and correspondences were publicly tied to their political reputation. It takes significant work to produce successful legislation, and much of that work can be productively and safely done without direct public oversight. This is the same reason an individual might value a private journal.

If you have a problem with your representative’s behavior, call their office and ask about it. They will certainly have a reason. Tell them if you disagree, and vote when you can.

Would you disagree that, given the choice, a solution to every problem is strictly better than a solution so only some problems?

A real treat for all the pilgrims on the Camino this year

Pronounced like “groom” makes for a nice analogy with slimming down the model size too.

This. My wife and I can hardly buy each other surprise gifts because the targeted advertising gives us away every time.

Can an industry insider confirm this tracking? Interesting take, thank you. I hadn’t considered the privacy implications of tipping.

In my research I take notes exactly as described here. I use plain-text files, one per week, with dated sections using markdown-ish notation where convenient. Display is never a goal; approximately 80-char column plaintext is the target format.

I agree with other commenters here that typing gives me more flexibility, in particular when writing arguments. I’ll format each point as a bullet and rearrange the list until I’m satisfied with the flow.

The notebook is essential for recovering tidbits learned along the way, e.g. what tricky steps did I need to get that one dependency to build. Weekly notepads are coarse enough to search by memory and contain enough context to get oriented quickly when going back several months.