Notices by Victorian Sparkgap Kitsune (enkiv2@niu.moe), page 2

Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Saturday, 21-Apr-2018 14:43:34 EDT Victorian Sparkgap Kitsune

EoE was an unexpected follow-up, in the same way Onegai Twins was unexpected given Onegai Teacher -- it's not a phone-in like Slayers Try or large parts of Saber Marionette J to X (which, if you managed to get through it, was ultimately salvagable). But, it also made a statement about its divergence.
Of course, the whole scheme was an elaborate ode to Ideon Be Invoked, and a way to make their influences more obvious to people who had missed them the first time.

In conversation Saturday, 21-Apr-2018 14:43:34 EDT from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Saturday, 21-Apr-2018 14:40:03 EDT Victorian Sparkgap Kitsune

It would be interesting to live in a world where Death & Rebirth and EoE never happened, and instead we got the second season that had originally been pitched, wherein EoTV was retconned immediately and instead the whole story is about Werewolf Asuka.
I think, in such a world, Evangelion would not be seen as a classic but instead one of those forgotten gems that were briefly extremely popular but doomed by weak follow-ups (like Saber Marionette J, Slayers, or Onegai Teacher).

In conversation Saturday, 21-Apr-2018 14:40:03 EDT from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Saturday, 21-Apr-2018 14:36:58 EDT Victorian Sparkgap Kitsune
- H. Faust
@hfaust
Yeah, that's sort of my metaphor. Though, maybe a better one would be the Christmas Special, since the Prequels are at least structurally, thematically, and continuity-wise connected to the original trilogy in a way that Rebuild 3 and 4 could not be.
(It's not just that. Evangelion is the Star Wars of anime, kinda: a huge deal, marketed all to hell, and people think it's a lot more original than it really is because its biggest influences are considred deep cuts.)

In conversation Saturday, 21-Apr-2018 14:36:58 EDT from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Saturday, 21-Apr-2018 10:02:22 EDT Victorian Sparkgap Kitsune

since Moore's law ended in 2005, more than 80% of programmers currently working have never worked under a situation where hardware performance increased better than linearly over time.

In conversation Saturday, 21-Apr-2018 10:02:22 EDT from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Friday, 20-Apr-2018 23:16:20 EDT Victorian Sparkgap Kitsune

a couple important points:
1) the number of professional programmers doubles every 5 years, so at any given time half of all pro developers have less than 5 years experience.
2) 1970 is the point at which brogrammers began to outnumber the first generation (who were largely serious engineers, & designed every worthwhile idea in CS).
3) most of that first generation retired around 1995.
this is why tech is like Scarfolk, collectively reliving the 70s every 5 years.

In conversation Friday, 20-Apr-2018 23:16:20 EDT from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Wednesday, 07-Mar-2018 07:15:00 EST Victorian Sparkgap Kitsune
- onreact
@onreact
I am familiar with how the term is used in the security community. It's a subset of its general use.
Your toot is the equivalent of "hacking isn't programming! hacking is breaking into banks! look it up!"

In conversation Wednesday, 07-Mar-2018 07:15:00 EST from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Tuesday, 06-Mar-2018 17:07:33 EST Victorian Sparkgap Kitsune
- Midcentury Modern Cockatrice
I'm moving my main to @enkiv2
If you're following me here & aren't following me there, and want to, please do.
This account will remain an alt centered on discussion of anime.

In conversation Tuesday, 06-Mar-2018 17:07:33 EST from niu.moe permalink
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Tuesday, 06-Mar-2018 16:55:39 EST Victorian Sparkgap Kitsune

Programming is Forgetting: Toward a New Hacker Ethic - All... http://opentranscripts.org/transcript/programming-forgetting-new-hacker-ethic/
In conversation Tuesday, 06-Mar-2018 16:55:39 EST from niu.moe permalink
Attachments
1. Invalid filename.
  
  Programming is Forgetting: Toward a New Hacker Ethic - Allison Parrish | Open Transcripts
  
  from Open Transcripts
  
  I wouldn't be surprised to find out that many of us here today like to see our work as a continuation of say the Tech Model Railroad Club or the Homebrew Computer Club, and certainly the terminology and the values of this conference, like open source for example, have their roots in that era. As a consequence it's easy to interpret any criticism of the hacker ethic—which is what I'm about to do—as a kind of assault.
Nate Cull (natecull@mastodon.social)'s status on Tuesday, 06-Mar-2018 15:54:47 EST Nate Cull

A quote from a 2015 Slashdot interview with the creator of C++'s Standard Template Library, Alexander Stepanov, popped up in my birdsite feed:
https://interviews.slashdot.org/story/15/01/19/159242/interviews-alexander-stepanov-and-daniel-e-rose-answer-your-questions
<< I am still convinced that Simula/C++/Java style inheritance is unsound.>>
<<One of the biggest changes since then has been the growth of caches. Cache misses are very costly, so locality of reference is much more important now. Node-based data structures, which have low locality of reference, make much less sense.>>

In conversation Tuesday, 06-Mar-2018 15:54:47 EST from mastodon.social permalink Repeated by enkiv2
Nate Cull (natecull@mastodon.social)'s status on Tuesday, 06-Mar-2018 15:56:30 EST Nate Cull
in reply to

I'm wondering particularly how, or if, Lisp cons cells (which are the maximally non-local kind of 'node structure') can be made to work better with caching, while still preserving the simplicity and generality that makes them attractive.
Especially across the Internet, where things like proxies and file systems count as one of those many layers of cache.

In conversation Tuesday, 06-Mar-2018 15:56:30 EST from mastodon.social permalink Repeated by enkiv2
Nate Cull (natecull@mastodon.social)'s status on Tuesday, 06-Mar-2018 15:59:20 EST Nate Cull

"Did you hear about the new version of COBOL? It's called ADD 1 TO COBOL."
-- C++ STL library designer Alexander Stepanov.

In conversation Tuesday, 06-Mar-2018 15:59:20 EST from mastodon.social permalink Repeated by enkiv2
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Tuesday, 06-Mar-2018 13:29:20 EST Victorian Sparkgap Kitsune

Tracing fake... https://blog.acolyer.org/2018/02/19/tracing-fake-news-footprints-characterizing-social-media-messages-by-how-they-propagate/
In conversation Tuesday, 06-Mar-2018 13:29:20 EST from niu.moe permalink
Attachments
1. Invalid filename.
  
  Tracing fake news footprints: characterizing social media messages by how they propagate
  
  By adriancolyer from the morning paper
  
  Tracing fake news footprints: characterizing social media messages by how they propagate Wu & Liu, WSDM’18
  
  This week we’ll be looking at some of the papers from WSDM’18. To kick things off I’ve chosen a paper tackling the problem of detecting fake news on social media. One of the challenges here is that fake news messages (the better ones anyway), are crafted to look just like real news. So classifying messages based on their content can be difficult. The big idea in ‘Tracking fake news footprints’ is that the way a message spreads through a network gives a strong indication of the kind of information it contains.
  
  A key driving force behind the diffusion of information is its spreaders. People tend to spread information that caters to their interests and/or fits their system of belief. Hence, similar messages usually lead to similar traces of information diffusion: they are more likely to be spread from similar sources, by similar people, and in similar sequences. Since the diffusion information is pervasively available on social networks, in this work, we aim to investigate how the traces of information diffusion in terms of spreaders can be exploited to categorize a message.
  
  In a demonstration of the power of this idea, the authors ignore the content of the message altogether in their current work, and still manage to classify fake news more accurately than previous state-of-the-art systems. Future work will look at adding content clues back into the mix to if the results can be improved even further.
  
  TraceMiner consists of two main phases. Since information spreads between nodes in the network, we need a way to represent a node. Instead of 1-of-n (one hot) encoding, in the first phase a lower-dimensionality embedding is learned to represent a node. In other words, we seek to capture the essential characteristics of a node (based on friendship and community memberships). In the second phase the diffusion trace of a message is modelled as a sequence of its spreader nodes. A sequence classifier built using LSTM-RNNs is used to model the sequence and its final output is aggregated using softmax to produce a predicted class label.
  
  The first step utilizes network structures to embed social media users into space of low dimensionality, which alleviates the data sparsity of utilizing social media users as features. The second step represents user sequences of information diffusion, which allows for the classification of propagation pathways.
  
  Let’s take a closer look at each of these two steps, and then we’ll wrap up by seeing how well the method works compared to previous systems.
  
  Learning an embedding for social media users
  
  Social networks of interest for fake news spreading tend to be those with the largest reach – i.e., they have a lot of users. Hence there is comparatively little information about the majority of users in any given trace.
  
  Just as most words appear infrequently, and a few words very frequently:
  
  So most users appear infrequently in information diffusion traces, and a small number of users appear much more frequently:
  
  (Both of these plots are log-log).
  
  [The user and word frequency plots] both follow a power-law distribution, which motivates us to embed users into low dimensional vectors, as how embedding vectors of words are used in natural language processing.
  
  Two state of the art approaches for user embedding in social graphs are LINE and DeepWalk. LINE models first and second degree proximity, while DeepWalk captures node proximity using a random walk (nodes sampled together in one random walk preserve similarity in the latent space). These both capture the microscopic structure of networks. There is also larger mesoscopic structural information such as social dimensions and community membership we would like to capture (as is done for example by SocDim).
  
  … the ideal embedding method should be able to capture both local proximity and community structures… we propose a principled framework that directly models both kinds of information.
  
  The model combines an adjacency matrix and a two auxiliary community matrices, one capturing community membership and the other capturing representations of communities themselves as embeddings. Full details are in section 3.2 of the paper.
  
  Diffusion traces and sequence modeling
  
  The topology of information spreading is a tree rooted with the initial spreader (or perhaps a forest). Dealing directly with this tree structure leads to a state space explosion though: with n nodes (users), there can be different trees according to Cayley’s formula.
  
  The problem can be simplified by flattening the trees into a linear sequence of spreaders. For example, indicating that the message originate with node 35 at time , was spread by node 12 at time , and so on. The only requirement is that for any two nodes i and j in the sequence if i comes before j then i spread the message before j did. This reduces the number of possible diffusions networks to (you know it’s a bad problem when we’re talking about reducing it to a factorial!).
  
  When the trees are flattened like this, some of the immediate causality is lost. For example, consider this tree and it’s flattened representation:
  
  The problem here is that the direct dependencies we’ve lost are important. “For example, the information flow from a controller account to botnet followers is a key signal in detecting crowdturfing.” They should still be close to each other in the sequence though. That’s something an RNN can cope with…
  
  We propose to use an RNN to sequentially accept each spreader of a message and recurrently project it into a latent space with the contextual information from previous spreaders in the sequence… In order to better encode the distant and separated dependencies, we further incorporate Long Short-Term Memory cells into the model, i.e., the LSTM-RNN.
  
  Now it’s likely that the account which first originated a message is a better predictor of its class than the last accounts to spread it. So the diffusion trace is sent through the LSTM-RNN in reverse so that the originator is seen as close as possible to when the final prediction is made.
  
  TraceMiner in action
  
  TraceMiner is evaluated alongside an SVM classifier trained on message content, and XGBoost fed with pre-processed content produced by the Stanford CoreNLP toolkit. “XGBoost presents the best results among all the content-based algorithms we tested.” Variants of TraceMiner that use only microscopic network structure for node embeddings (i.e., DeepWalk and LINE) are also tested, to assess the impact of the community embeddings. Experiments are run on two Twitter datasets:
  
  With the first dataset the challenge is to determine the news category, which is one of business; science and technology; entertainment; and medical. (On this task, you would expect the message-content based approaches to do well). The second dataset contains a 50:50 mix of genuine news and fake news, and the task is to tell them apart.
  
  For the first news categorisation task, a variety of models are tested, using differing percentages of the training data. Two different accuracy measures are reported: Macro-F1 is just the average F1 score across the four categories, and Micro-F1 is the harmonic mean of the precision and recall scores. (The point of the Micro-F1 score is to be less sensitive to class imbalances).
  
  On the Micro-F1 measure, TraceMiner does best across the board, and outperforms all other approaches on the Macro-F1 score until the amount of data used for training goes above the 80% threshold. XGBoost (which looks at message content) does best on this measure at the 80% mark.
  
  With fake news, as we would hope, the advantage of the TraceMiner approach is even more distinct (based on the hypothesis that content is less distinguishing between real and fake news):
  
  Unlike posts related to news where the content information is more self-explanatory, content of posts about fake news is less descriptive. Intentional spreaders of fake news may manipulate the content to make it look more similar to non-rumor information. Hence, TraceMiner can be useful for many emerging tasks in social media where adversarial attacks are present, such as detecting rumors and crowdturfing.
  
  Now, in the case of fake news it is perhaps of some utility to be able to say after a piece of news has spread that “yes, that was fake news.” But by then it will have done it’s work and exposed its message to many people. So it’s notable in the table above that TraceMiner still performs well even with little training data.
  
  … optimal performance with very little training information is of crucial significance for tasks which emphasise earliness. For example, detecting fake news at an early stage is way more meaningful that detecting it when 90% of its information is known.
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Tuesday, 06-Mar-2018 13:27:59 EST Victorian Sparkgap Kitsune

As we may think | the morning paper https://blog.acolyer.org/2018/02/23/as-we-may-think/
In conversation Tuesday, 06-Mar-2018 13:27:59 EST from niu.moe permalink
Attachments
1. File without filename could not get a thumbnail source.
  
  As we may think
  
  By adriancolyer from the morning paper
  As we may think Vannevar Bush, The Atlantic, 1945
  
  To close out the week, here’s another selection from the ‘Great moments in computing’ list – and it’s a true classic. Bush’s article was written in 1945 when he was Director of the Office of Scientific Research and Development coordinating the work of about six thousand scientists in the war effort. With the war coming to an end, the central question Bush addresses is what scientists should now turn their attention too.
  
  There is a growing mountain of research. But there is increased evidence that we are being bogged down today as specialization extends. The investigator is staggered by the findings and conclusions of thousands of other workers — conclusions which he cannot find time to grasp, much less to remember, as they appear…. Professionally our methods of transmitting and reviewing the results of research are generations old and by now are totally inadequate for their purpose… The summation of human experience is being expanded at a prodigious rate, and the means we use for threading through the consequent maze to the momentarily important item is the same as was used in the days of square-rigged ships.
  
  We need better approaches to tap into this body of knowledge. How can machines help us in our thinking processes? To be useful to science a record must be continuously extended, it must be stored, and above all it must be consulted.
  
  When it comes to recording information, Bush envisions a small camera attached to the forehead, by which a scientist moving about the laboratory of field can capture a picture of anything of interest. And whereas today (1945) we use a pencil or a typewriter, ‘will the author of the future cease writing by hand or typewriter and talk directly to the record?’.
  
  One can now picture a future investigator in his laboratory. His hands are free, and he is not anchored. As he moves about and observes, he photographs and comments. Time is automatically recorded to tie the two records together… As he ponders over his notes in the evening, he again talks his comments into the record. His typed record, as well as his photographs, may both be in miniature, so that he projects them for examination.
  
  To leave time for high-value thinking activities, machines should automate as many repetitive or rule-based tasks as possible:
  
  Whenever logical processes of thought are employed — that is, whenever thought for a time runs along an accepted groove — there is an opportunity for the machine.
  
  For example, “it is readily possible to construct a machine which will manipulate premises in accordance with formal logic.” Beyond the logic of the mathematician, lies the application of logic in everyday affairs, ”we may someday click off arguments on a machine with the same assurance that we now enter sales on a cash register.”
  
  So much for the manipulation of ideas and their insertion into the record. Thus far we seem to be worse off than before — for we can enormously extend the record; yet even it its present bulk we can hardly consult it. This is a much larger matter than merely the extraction of data for the purposes of scientific research; it involves the entire process by which man profits by his inheritance of acquired knowledge.
  
  One of the deep problems that Bush highlights is that selection of information is primarily via indices, and yet the human mind works by association. We create trails of thought. And this leads us to the famous memex…
  
  The memex
  
  Consider a future device for individual use, which is a sort of mechanized private file and library. It needs a name, and, to coin one at random, “memex” will do. A memex is a device in which an individual stores all his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility. It is an enlarged intimate supplement to his memory.
  
  You can buy content (on microfilm) ready for loading into the memex — books, pictures, periodicals, newspapers and so on. Business correspondence can also be saved. For you own content you can place longhand notes, photographs, memoranda etc., on top of a special platen on the desk, whereby pressing a special lever causes it to be photographed and placed onto the next blank space in a section of the memex microfilm.
  
  Fast movement forward and backward when reviewing content (e.g., a book) is via special levers which allow movement forward and backward by 10 pages at a time with a short throw of the lever, and 100 pages with a long throw. With a dedicated button you can go immediately to the first page of the memex index. The memex desk has several positions to which content can be projected, thus the user can leave one item in position while calling up another.
  
  He can add marginal notes and comments, taking advantage of one possible type of dry photography, and it could even be arranged so that he can do this by a stylus scheme…
  
  It is associative indexing though, that is the essential feature of the memex, “the process of tying two items together is the important thing.” Bush describes a hypertext like mechanism at this point, but most interesting from my perspective is his emphasis on a trail as a fundamental unit — something we largely seem to have lost today.
  
  When the user is building a trail, he names it, inserts the name in his code book, and taps it out on his keyboard… when numerous items have been thus joined together to form a trail, they can be reviewed in turn, rapidly or slowly, by deflecting a lever like that used for turning the pages of a book. It is exactly as though the physical items had been gathered together from widely separated sources and bound together to form a new book. It is more than this, for any item can be joined into numerous trails.
  
  Trails can be shared with others, for insertion into their own memexes. What kind of things might we build with such trails:
  
  “Wholly new forms of encyclopedia will appear, ready made with a mesh of associative trails running through them, ready to be dropped into the mesh and there amplified.”
  
  Lawyers have at their touch, “the associated opinions and decisions of their whole experience, and of the experience of friends and authorities.”
  
  Patent attorneys have details of millions of issued patents, with familiar trails to every point of their client’s interest.
  
  Physicians can run rapidly through case histories, with side references to the classics for the pertinent anatomy and histology.
  
  Chemists have all the chemical literature before them, with trails following the analogies of compounds, and side trails to their physical and chemical behaviour.
  
  There is a new profession of trail blazers, those who find delight in the task of establishing useful trails through the enormous mass of the common record. The inheritance from the master becomes not only his additions to the world’s record, but for his disciples the entire scaffolding by which they were erected.
  
  Documents and links we have aplenty. But where are our trails?
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Tuesday, 06-Mar-2018 13:22:16 EST Victorian Sparkgap Kitsune

Counterintuitive Properties of High Dimensional Space... https://marckhoury.github.io/counterintuitive-properties-of-high-dimensional-space/
In conversation Tuesday, 06-Mar-2018 13:22:16 EST from niu.moe permalink
Attachments
1. File without filename could not get a thumbnail source.
  
  Counterintuitive Properties of High Dimensional Space
  
  from Lost in Spacetime
Vertigo (vertigo@mastodon.social)'s status on Tuesday, 06-Mar-2018 13:08:27 EST Vertigo
- Victorian Sparkgap Kitsune
@enkiv2 from the video, "..., so this guy already has a program which is five to seven CHARACTERS long, and he hasn't gotten it to produce a result yet."
(Emphasis mine.)
I love that. Characters. Not statements, not words -- characters. LOL!
As a Forth junkie, we're happy if we can yield a colon definition that is small enough to fit on a single line (8 words on average).

In conversation Tuesday, 06-Mar-2018 13:08:27 EST from mastodon.social permalink Repeated by enkiv2
Victorian Sparkgap Kitsune (enkiv2@niu.moe)'s status on Tuesday, 06-Mar-2018 12:42:19 EST Victorian Sparkgap Kitsune
- Vertigo
@vertigo thanks!

In conversation Tuesday, 06-Mar-2018 12:42:19 EST from niu.moe permalink
Vertigo (vertigo@mastodon.social)'s status on Tuesday, 06-Mar-2018 12:38:21 EST Vertigo
- Victorian Sparkgap Kitsune
@enkiv2 Here's a link to the video corresponding to the talk this came from. https://dyalog.tv/Dyalog17/?v=9xCJ3BCIudI

In conversation Tuesday, 06-Mar-2018 12:38:21 EST from mastodon.social permalink Repeated by enkiv2
Woozle Hypertwin (woozle@toot.cat)'s status on Tuesday, 06-Mar-2018 08:32:47 EST Woozle Hypertwin
in reply to

LB: no confirmation yet as to whether this is genuine or someone has hacked the account... but it would just too perfectly epitomize the "quarterly profits projection" frame of mind if they put all that effort into customizing Masto, tacking technocapitalist stuff on, and promoting it -- but then decided to shut down because revenue wasn't as much as they'd hoped.

In conversation Tuesday, 06-Mar-2018 08:32:47 EST from toot.cat permalink Repeated by enkiv2
lute bird (murkrow@witches.town)'s status on Tuesday, 06-Mar-2018 08:29:27 EST lute bird
in reply to
- Hiveway
@hiveway *presses f to pay respects*

In conversation Tuesday, 06-Mar-2018 08:29:27 EST from witches.town permalink Repeated by enkiv2
Hiveway (hiveway@hiveway.net)'s status on Tuesday, 06-Mar-2018 08:14:26 EST Hiveway

Dear all, this instance will be shutdown tomorrow. Thank you all for being here. Over and out.

In conversation Tuesday, 06-Mar-2018 08:14:26 EST from hiveway.net permalink Repeated by enkiv2

After
Before

Public

Notices by Victorian Sparkgap Kitsune (enkiv2@niu.moe), page 2