stefanpie

I remember the original twitter thread on the topic but haven't heard much about it since then. It would be interesting to see a nice arxiv submissions with some initial experiments across common tasks and architectures. Would also be interested in hardware efficient versions for fixed point data.

Here is some background and more thoughts/references on this on this topic: https://github.com/brohrer/sharpened-cosine-similarity

Link to the relevant journal article: https://ieeexplore.ieee.org/document/9714188

As a novice Linux (Ubuntu) user, I am curious as to how these new changes for rand will speed up a Linux distro overall or if there is even a way to measure this. It seems that, in addition to an overdue replacement SHA-1, the new code is a significant increase in performance. I would think that there are a lot of places where random used in the OS and this update would translate into performance improves in those other places. I'm not an expert and would love to head any thoughts on this.

I actually ended up doing some variation of this for scripts I wore for my labs. The space is actually quite small and it's not hard to brute force this problem. Once you get to more open ended designs and larger parameter spaces (like filters) it becomes a bit more challenging to brute force.

I'll take a look at this, thank you for the node. I believe I'm using mathjax currently but I may have set set some options correctly. I also need to go in and verify a11ty as well.

After some quick searching (https://www.princeton.edu/~aaa/Public/Teaching/ORF523/S16/OR...), I don't think so. It looks like for robust optimization problem you don't know for certain what the objective function is or have uncertain data measurements for the objective function which I don't think is the case here. I could be completely wrong though.

From my understating, if you want to only able to only pick form a discrete set of components for certain variables, you are essentially transforming the problem into a Mixed-Integer Non-Linear Optimization problem (MINLP). I tried to find and easy way to do this with some python optimization libraries but they always needed other packages which were hard to install on Windows. So my solution was to "relax" the discrete constraints by using a continuous approximation of the minimum function when looking at the error from the nearness comment value that is possible. This also lets you assign weighting to have a tradeoff between having realistic comment values vs. having lower error for your main objective.

Oh interesting application. I have taken a class in microelectronics physics (doping, electron holes, bandgaps, mosfets, bjts) and it was definitely just focusing on simpler approximations and first and second order effects only.

But you are correct in that after my second or third electronics lab I realized that even standard component values vary to much that this approach is a little overkill and non-practical in a lot of cases. However there is a way to bake in discrete component values into the continuous optimization problems by using a continuous/"soft"/differentiable approximation of the min function to make "pockets" of optimal regions close to component values part of a set (for example e series values). I plan to do more writeups exploring this idea as well as looking at more complex applications such as higher order filters as you mentioned (my op-amp filter labs are what motivate me to look into this as I wasted so many hours in open lab trying to get my component values right).

Should be back to normal again, apologies for the issues, web dev is not my strong point

You are totally correct, this is only a typo in latex which I have just pushed a fix for. Thank for you the note.

I don't know how well this relates, but the topic of logistic growth (exp. growth with a limit) is common in undergraduate differential equations classes. I guess it can work well for simpler models.

https://en.wikipedia.org/wiki/Logistic_function#Modeling_ear...

I would say if you have the money for it, the Icebreaker FPGA dev board is really nice and integrates with the open source tools well. When prototyping I either use Verilog + Iverilog for simulation + yosys and icestrom tools for building, or more recently, I've been messing around with the active nmigen fork for python and doing design and simulation all in there which also links to the other tools for build as well. Honestly what software you like just depends on what you find best.

Robert Erdmann also gave a keynote talk at PyCon US 2021 about how they used Python for most of digitization and processing of the Night Watch (aka Operation Night Watch). Fascinating talk with tones of technical details.

Link: https://www.youtube.com/watch?v=z_hm5oX7ZlE&list=PL2Uw4_HvXq...

I'm surprised no comments here are in praise of Eleventy (11ty). As someone who works on more hardware-oriented stuff and not web dev, 11ty was the simplest to understand static site generator I've tried. I started out with Hugo, but there was too much going on for me to figure out how it works. Most tutorials or google searches ended up just pointing to "use this theme and change/add these files to add new stuff". 11ty is so dead simple in comparison. You just say, "hey, I have these files using these templates," and it just spits them out rendered. If you want to group certain data, you just label them using front matter and put them in a folder. And if you want to put content from one file into templates, you just specify what template file you want to use. I think it's the perfect balance between simplicity and customizability. Being able to write posts and markdown and throw together a simple HTML template for the base layout has been a lovely workflow to use.

I am trying to figure out how these work because this seems interesting to me. My understanding is that the "game books" tell you the rules and the scenarios. But how do you play the games: software, on a table like board games?

I'm familiar with graphs but not graph subsampling. What is graph subsampling used for and what are some applications that can use it?

Sounds like Home Assistant may fit your usecase well if you are willing to do some small initial configuration. https://www.home-assistant.io/

For personal projects, Kanban has been a very good way to stay focused on incrementally building features for a minimal prototype of an idea. With free tools like Trello and GitHub projects it seems more for for personal use.

Just took a radars course and we definitely spent a whole section on just atmospheric effects / defraction. My only guess about how you could maybe lower the error is by monitoring and measuring the condition of the atmophere and building a model that can use that data to apply corrections and reduce error to your image.

I think almost every FPGA vendor development tool has integrated tcl support which is basically the standard for scripting in that industry. Vender lock-in exasperates the trend.

Do all Georgia Tech students have access to the Safari? (coming from a fellow GT student)

Look like the data comes from a NASA satalite that tracks measurements for space weather and related scientific missions. Monitoring/forecasting space weather, especially solar effects and impacts, is important for detecting solar flares and ejections that can disrupt electronics in space and on the ground.

Very much like the plot of the black Mirror episode "Hated in the Nation"

It is, and the research applying ML in this area is starting to ramp up. For example, last year I worked on a project using ML to identify tornado vortex signatures in Doppler weather radar scans. It also turned out that a couple other groups published similar research at the same time. I would say to expect to be much more growth with ML in meteorology hoping it will all eventuality be applied in the field.

I have found Matlab as a great tool to write scrips to do data exploration and manipulation; then later on I can easily reimplement the same workflows or algorithms into something like a python notebook and also have the added benefit of more flexible and open frameworks (for stuff like ML with keras and tensorflow).

A quick google search yielded this:

http://excamera.com/sphinx/fpga-cpu.html

QR codes also will always work in grayscale.

Then isn't that more analogous to Early Action which is nonbinding yet still being able to apply and hear your decision earlier.

I wanted to get into FPGA programming recently after feeling confident mastering my microcontroller skills and discovered that complex programmable logic devices (CPDLs) can be a stepping stone to full on FPGA prototyping. They somewhat of a similar high-level functional design and there are some nice guides online to get started on some projects.

I immediately thought of the "Hated in the Nation" episode of Black Mirror.