HNHacker News
TopNewBestAskShowJobs

w_t_payne

1,510 karma · joined May 23, 2012

blog.wtpayne.info
submissionscomments
w_t_payne··on Three authors retract study on hydroxychloroquine and chloroquine
Since I take it daily for my Rheumatoid Arthritis -- a source of more than a small amount of concern.
w_t_payne··on Researcher questions years of his own work with a reexamination of fMRI data
We still have a whole heap of mystery about us - especially where the brain is involved. :-)
w_t_payne··on Ask HN: Hacker tips to maintain a good romantic relationship?
A romantic interest isn't really compatible with deep focus work, IMHO, so I'd argue that you're better off not attempting to maintain one in the first place.
w_t_payne··on Trump's executive order targets political bias at Twitter and Facebook: draft
On the face of it this seems to be a dangerous step towards the industry being coopted by the lunatic conspiratorial fringes and forced to spread their dangerous nonsense against our will. Not a game I'd be happy to play.
w_t_payne··on Ask HN: Why isn’t visual programming a bigger thing?
I was aware of OpenModelica, but hadn't looked at the project in any great depth. I think it's a lot more ambitious than my project, which is restricted (at present) to a single model of computation.
w_t_payne··on Ask HN: Why isn’t visual programming a bigger thing?
I'd love to have a chat with you about visual programming - I've got my own project that I'm (glacially slowly) pushing forward in my spare time, and although it's not got an interactive UI yet -- that's very much the end goal.
w_t_payne··on Ask HN: Why isn’t visual programming a bigger thing?
My personal mission is to produce a version-control-friendly and open-source LabView/Simulink killer. Visual display for the managers, textual editing to save the poor engineers who have to program the thing from mouse-hand RSI. Do the whole thing in Python/C to make it both hip and fast.
w_t_payne··on Ask HN: Why isn’t visual programming a bigger thing?
I'd love to see that as well. I might try to give it a go for Python code - try to generate the control flow graph, data dependency graph and syntax tree, and find a way of flipping between them.
w_t_payne··on Ask HN: Why isn’t visual programming a bigger thing?
I'm taking a very very similar approach -- starting off with a computational model driven by an easy-to-diff data structure that can be serialised to YAML/JSON/XML and then looking for ways of building a UI on top of that. I'd love to share experiences with you -- NDA permitting, of course.
w_t_payne··on Ask HN: Why isn’t visual programming a bigger thing?
Visual programming has the potential to be really useful ... but I haven't seen anyone get the UX quite right just yet (although a few have come close).

I spent a few years programming mainly with Simulink, and more recently have experimented a little with some graph-based game engine UIs. (Unity shader graph etc..)

Now, as far as Simulink is concerned, I feel that it was (and possibly still is) an ergonomic disaster zone. There's just too much mousing around and adjusting routing between nodes. Also, merging is really difficult because of the save file format. This is a significant problem for many engineering organisations.

For any visual programming tool, the value of a visual graph diminishes as the graph becomes more complex. It is most valuable when it is kept simple and illustrates e.g. high level components only.

Now ... none of these problems is insurmountable, and having a relatively simple high-level graph to show the relationship between the major components of a system is an incredibly valuable communications tool -- but users do need to resist the temptation to use the graph for everything down to every if statement and for loop. These things are best used to explain the high level relationship between major system components and the overall flow of data through the system. Fine grained algorithms are better represented textually.

So, over the years I've developed a configuration-driven framework that's designed to (eventually) be driven from a visual UI.

Computationally, the framework implements a Kahn process network. I.e. it's a dataflow model, where the nodes represent sequential processes, and the edges have queue semantics. (The queue implementation can be replaced with implicitly synchronised shared memory in some situations, so it can be quite efficient).

This enables me to e.g. implement intelligent cacheing of data flows, support record-and-replay style debugging, co-simulation, and automatic component-test generation.

The nodes can be arbitrary unix-philosophy binaries, Python functions, or native shared libraries. (Eventually I plan to support deploying nodes to FPGA as well), and because of the Kahn process network semantics, behave in the same way irrespective of how the nodes are distributed across computers on the network. This makes it an ideal rapid-prototyping tool for quickly integrating existing software components together. E.g. throwing machine learning components together that have been written in different languages or using different frameworks. It's a bit like ROS in this regard.

The dataflow graph itself is represented by a set of YAML files (or data structures that can be generated/modified at runtime), with different aspects (connectivity, node definitions, layout) separated to make textual diffs and merges easy to understand, and enabling teams to work more effectively together.

Also, because the graph is explicitly represented as a design-time data structure rather than being a runtime construct, it's easy to use GraphViz to generate diagrams, allowing you to have documentation that's correct by definition without spending ages adjusting untidy connections and layouts. (Although I do want to spend some time improving the layout algorithms and providing some mechanism for layout hinting).

Right now, you can only generate the visual graph from the textual description. In future I want to make it possible to edit the graph visually, so you can (for example), drag and drop compute nodes onto different computers really easily; stop, rewind and replay history, and simulate the effects of e.g. network contention and experiment with moving nodes around the available hardware to optimise performance.

I'm also experimenting with hybrid symbolic/statistical AI techniques that use the data flow graph to help with fault finding, automatic safety-case generation etc.. etc..

w_t_payne··on Ask HN: What's your quarantine side project?
A distributed simulation framework for Python based on Kahn process networks, intended to provide support for design automation based on a novel hybrid of statistical techniques (including DNNs) and rules-based AI.
w_t_payne··on ImageNet-trained CNNs are biased towards texture (2019)
I don't think that's what's happening here.
w_t_payne··on Study: No evidence of efficacy of hydroxychloroquine in hospitalized patients
Given that I need hydroxychloroquine to treat my rheumatoid arthritis, I'd like this hypothesis to be either proven or disproven as rapidly and as unambiguously as possible.
w_t_payne··on Oxford Medical model says defeating covid19 requires digital contact tracing app
GeoLocal contact tracing - Walk a bipartite graph - one set of nodes formed from mobile phone (or App) unique identifiers, the other set of nodes formed from location/time identifiers - such as IP addresses + day, or lat/lon + day. It won't give you who sat next to who on the bus, but it's better than nothing.
w_t_payne··on Germany tries to stop U.S. from luring away firm seeking coronavirus vaccine
Trust is a key commodity in times of crisis. A commodity that, sadly, has been squandered all to readily in recent years.
w_t_payne··on Generative Teaching Networks: Accelerating Neural Architecture Search
I really like the idea of optimising the 'direct' training data, and wonder how it would interact with the use of synthetic data as the 'indirect' training data. Or perhaps some sort of restriction on the (optimised) 'direct' training data as a form of regularisation. Lots of potential ideas to explore here.
w_t_payne··on Google users in UK to lose EU data protection – sources
This move will make it considerably easier for influence shops to shape public opinion in the UK and could help open up e.g. healthcare markets which are currently closed.
w_t_payne··on Do Software Engineers inflate the reported hours they claim to work?
A few years ago (2014ish) I would start the work week pulling 7am to 10:30pm - approx 15 hour days -- but couldn't sustain that for more than a couple of days, so by Friday my hours would drop to 10-11 hours per day, and I would spend pretty much the entire weekend asleep - so significantly less than the 90h claimed and much closer to 60h. After a couple of years of this, I was in the middle of an emotional breakdown with rapidly encroaching early-onset rheumatoid arthritis - my body couldn't handle it any more. So ... while I don't think it's totally impossible - it's also a long way from being a healthy lifestyle. (Needless to say I try to avoid this sort of thing these days).
w_t_payne··on Tesla's self driving algorithm's overlay [video]
Vehicle-to-X (A generalisation of vehicle-to-vehicle communication) https://en.wikipedia.org/wiki/Vehicle-to-everything
w_t_payne··on Managing my personal knowledge base
So, renaming is done manually (forces me to look at the PDF). I use preview on OSX, or any viewer other than Acrobat Reader on Windows and Linux (Acrobat Reader locks the file preventing renames). Several Python libraries exist for extracting metadata. I'm trying a couple of different approaches. At the moment I just take the title of the PDF and do a search on Google Scholar using the scholarly python library -- but this is really very suboptimal and I want to replace it with something faster and more robust.
w_t_payne··on How Germany is able to run the world’s second largest export economy
Competing with China, Germany and the US is possible -- but you have to be deliberate, disciplined and very very very organised.
w_t_payne··on Managing my personal knowledge base
I can have multiple copies of the same document in the system. My notes are associated with documents via md5 hash, so it'll link them with all copies that are present. At some point I'll get the script to automatically hardlink or symlink duplicates. To be honest though, trying to decide which location best fits a document is quite a good way to engage with the document content -- so even though decisions are often suboptimal, it isn't really the final hierarchy which is the product here, it's what goes on in my brain during the process of trying to organize the papers. As far as the NLP libs question is concerned -- so far this is just an excuse for me to play with SpaCy .... I'm still at quite an early stage and haven't made much progress (my background is more machine vision than NLP - which I haven't touched since I was an undergrad 20 years ago).
w_t_payne··on Managing my personal knowledge base
I've got a hybrid system.

Academic papers, standards documents, white papers etc... get downloaded as PDF and renamed according to a (human-readable) standard scheme that I use (year, title, [author(s)], [paper-type]), and stored in a 'meaningful' filesystem hierarchy. Podcasts and some videos are also stored in the same hierarchy.

Documentation for the software libraries and APIs that I use is downloaded from readthedocs (where possible) and stored in a parallel system that takes account of versioning. (So I can concurrently store differently versioned copies of the documentation for a single library).

I have a simple python script that iterates through my directory hierarchy and produces a sqlite database and a couple of xlsx files with a row-per document (one spreadsheet for reporting document-management metadata and another that allows me to assign labels and write precis notes). The script also extracts the content of the PDFs as plain-text and feeds some NLP tools that I'm playing with.

I use the spreadsheets to keep notes on the files, and the act of manually renaming and sorting the PDFs into the 'right' place in the hierarchy helps me to understand what's in them and remember what I've got. (I'm constantly reorganising the hierarchy as my understanding develops and evolves. My python script keeps everything -- notes and documents and other metadata - in sync).

So far this has scaled OK to around 26,000 PDFs.

w_t_payne··on Gitlab considers not hiring SREs and Support Engineers in China and Russia
This is where it is important to get back to first principles, and making distinctions based on nationality is less helpful than making distinctions based on rules-oriented-organisations vs. people-oriented organisations. The primary distinguishing characteristic which we need to pay attention to is rule-of-law and the power-predominance of rules-oriented-organizations across the political and economic system. This is something which should in principle be apolitical and possible to evaluate empirically.
w_t_payne··on Hydra: A framework that simplifies development of complex applications
So ... what does it do, exactly?
w_t_payne··on Pedestrian detection systems don’t work well, AAA finds
I have former colleagues working on some of these systems, and they tell me some pretty scary horror stories of the simple and obvious ways in which they fail. Some of the stories that I have heard just blow my mind - and seem all-too-reminiscent of the VW emissions-test defeat-device scandal.
w_t_payne··on Ask HN: What happens after no-deal Brexit?
I am quite close to a couple of people actually working on Brexit ... and while of course they won't tell me any specifics, the message that I'm getting from the body language, eye-rolls and exasperated sighs is ... pretty bleak.
w_t_payne··on Conventional Commits: A specification for structured commit messages
I have a system that creates commit messages automatically. The commit messages themselves are YAML so that they can contain various bits of metadata - current task id, timestamps for oldest/newest known builds associated with that task etc...
w_t_payne··on We can’t trust AI systems built on deep learning alone
Differentiable SAT solvers is a thing.
w_t_payne··on The Spinner
Like any technology it can be used for good or evil ... but it has to be said ... anger is a helluva drug.
← PreviousPage 6 of 32Next →