They are actually pretty simple. Essentially you generate the Dijkstra values for an undirected graph (this can be a grid, navmesh, etc), then you create directed edges pointing from high values to low values. So a grid space of value 8 will point to its neighbors with values less than 8, etc.
All an agent has to do is query their current spot in the graph and it will return a vector that leads them to the next lowest cost. This is useful if you have lots of agents going to the same location.
The description of this video has a lot of good resources. I made it when I was a much much worse programmer though so I wouldn't bother actually watching the video lol.
I am not sure why you are being downvoted. I have been doing a project with Homogeneous Transform Matrices recently and also assumed that this was "obvious" since the shear diagonal lives in the rotation part of the matrix. Still, it's a very neat trick and I appreciate the OP for posting it!
> I beg to differ. It was radar-guided anti-aircraft guns
The 8.8cm german flak gun WAS a radar-guided AA gun.
> and proximity fuzes for artillery shells
The axis may not have had proximity fuzes, but the 8.8cm flak was so accurate that during a statistical analysis it was determined that the part of the gun that introduced the most accuracy were the timed fuzes. They switched to shells that detonated on impact and actually ended up scoring more kills than before.
The actual piece of equipment that made this thing so accurate was the fire control computer, the Kommandogerat 40.
Here is a Google Drive link to some scans I made in college of the book written by the designer. It explains all the calculations that the analog computer is doing.
> The 8.8cm german flak gun WAS a radar-guided AA gun.
The Kommandogerat 40 was a rangefinder/computer, not a radar. It had the capability of being radar-guided, but was not necessarily radar-guided by itself.
For the Kommandogerat 40 you needed a direct line-of-sight (they were still used in overcast conditions though) but it would compensate for the velocity vector of the incoming targets, altitude, wind, and would relay fire controls to all the flak pieces in the battery. Not as nice as a radar, but it was able to be fielded in mass quantities, and I think it deserves its place as 2nd most devastating weapon of WWII. Especially since towards the end of the war, the Germans were the ones playing defense. It made the strategic bombing campaigns of the allies very painful. The RAF bomber crews had a 44% death rate. That's deaths, not casualties.
It also excelled in direct fire roles against ground targets.
I read that it took 16,000 88 flak shells to bring down one bomber. I infer the success of the flak was due simply to quantity.
I've also read numerous times that the searchlights would find and "pin" a bomber so that the crews could dial in the coordinates for firing. Wouldn't need that with radar guided guns.
I well know the death rate for B-17 crews - my father flew 32 missions in 1944 in one. The casualty rate was 80% (killed, wounded, POW). He told me that the way he dealt with it was simply accept that since he was going to die, he'd do the best job he could. I have the letter he wrote his father, dryly saying he'd completed his missions and would be coming home.
The Bismarck was well equipped with anti-aircraft batteries, which failed to stop the torpedo attack by slow-moving, obsolete biplanes. The Japanese navy didn't have much success with anti-aircraft batteries, either.
> I read that it took 16,000 88 flak shells to bring down one bomber. I infer the success of the flak was due simply to quantity.
I am not trying to be pedantic, just wondering out loud. 16,000 shells per bomber sounds like a lot, but I am sure you are familiar with the statistic that 250,000 small-arms rounds were fired for every insurgent killed in Afghanistan, and 25,000 rounds fired per kill in WWII. I can't find an actual hard source on the ratio of bullets per kill though. Bombers obviously cant hide behind cover, but I wonder if that 16,000 number isn't as unreasonable as it seems.
> I well know the death rate for B-17 crews - my father flew 32 missions in 1944 in one.
He sounds like an amazing man. I get queasy in light turbulence. I can't imagine the stress and horror of even a single bombing run. That's more than anyone should ever have to go through.
He said many of the men cracked under the strain. Some would deal with it by drinking themselves to oblivion.
> He sounds like an amazing man.
He was, but I'm rather biased :-)
He told me once he wouldn't trade that experience for anything, and wouldn't do it again for anything. He also said that when he felt down, he'd think of the men he knew who didn't get the chance to live, and would fix his attitude.
> 250,000 small-arms rounds were fired for every insurgent killed in Afghanistan
To be fair, only a small part of that is aimed. Much of it is "covering fire", the intent of which is to cause the enemy to hide while your guys advance. There are also soldiers who do not wish to kill, and do not aim their shots.
The larger point is that if the accuracy of the flak were doubled with radar guidance, it would have been impossible for the 8th AF to continue the mass bombing campaign. If 5% are lost on a mission (and that did happen now and then) repeatedly, then it ain't long before there isn't an Air Corps left.
Build up my savings to the point where I can live comfortably for 2+ years without a paycheck, or 4+ on a tight budget.
DCA into ETFs with the leftovers.
Buy a little bit more non-perishable food than I need every time I go to the store. Extra bag of beans and rice here and there. Also eating unhealthier than usual since I figure if there's a food shortage or energy shortage during the winter I'd rather be 15lbs overweight.
Continue to learn skills that I think will be valuable in the future/will get me ahead of the average person in my field. I was a child in 2008 so I have never experienced a bad labor market, but I figure I just need to be more employable than average (if I am mistaken then that's what the 2+ years of savings are for). There is always money to be made and people will want to hire those with skills that can make them money. If tech is in a bubble then just being able to make CRUD apps wont land six figure salaries anymore. I'll need a good reason for someone to hire me.
My weight fluctuates pretty normally depending on my exercise level. When I'm actively lifting weights I am in a calorie surplus and usually end up a bit overweight. Losing weight is fairly easy. You just count calories, avoid sugars, and eat high-fiber foods at the start of the meal to feel more full for longer.
I recognize that 'gain weight for the winter famine' is a bit extreme but I'm just willing to gain 5 extra lbs than I normally would.
Human brains are funny. Obviously the absurd part of all of this is people (literally) building their houses on the edge of the ocean and expecting this to not happen.
However because it's all we know, it seems that the real absurdity would be the government designating these sites as 'no-build' zones-- or at least FEMA washing their hands of any damage that should happen to these sites.
This should be a wakeup call that many aspects of our current way of life are unsustainable, not that it's just standard operating procedure to rebuild from the ground up every 20 years only for everything to be washed away again.
I would wager a majority of the final costs are from damages to houses not "on the edge of the ocean". Hurricane damages stretch for miles inland. Heck pick any east coast state and check out the areas covered by flood warnings and watches. They're hours from the coast in places. If it was only the houses on the edge of the ocean, this would be relatively cheap.
That's a good point, I decided to do some napkin math:
Sanibel Island has 7,821 housing units at an average density of 454.6 per square mile.
Median sales price of the homes in 2021 was $1.64M, average price was similar.
So a rough approximation of the value of the homes on the island alone is $12Bn. I am not sure how you would translate that into damage costs, but in a scenario of total destruction then $12B would be lost.
I have no real evidence other than visiting Ft Myers and Naples a few times, but I can tell you there is (was) a lot of very valuable property right next to the ocean. I don't doubt that if you look at the total distribution of damages that there is a long tail of less severe damage as you go inland, but the conservative estimates from my napkin math shows that it's likely a significant number (over ten billion) occurred just from properties on the small stretch of coastline.
Insurance isn't going to pay out on value though, it's going to pay out on replacement costs. A home that's $5M because it sits on the waterfront would be $750K 5 miles inland, but they will cost roughly the same to replace.
I sincerely doubt the cost to replace the first home is 6x the second. And then there's this one (https://www.trulia.com/p/fl/sebring/2124-gardenview-rd-sebri...), which is about as far from the coast as you can reasonably be in that part of florida, going for a mere $240k, and I doubt it's significantly more or less to replace than either of the previous two.
Fgemm is based on the initials sgemm or dgemm, based on gemm, GEneral Matrix-Matrix multiplication. gemv is GEneral Matrix-Vector multiplication. So it's meaningless, except of course for my target audience, who gets spinal suffering from typing those letters out in that order, everybody knows what it means. fgemm, if you google it, gets a lot of hits. Supposedly it means like a modular form of gemm, but there's also, among the hits, among which I am not counted despite owning the domain, the intended initials.
Fast GEneral Matrix-Matrix Multiplication.
Faster GEneral Matrix-Matrix Multiplication.
Fastest GEneral Matrix-Matrix Multiplication.
One of those.
I'm not sure which yet. Probably Fastest. That's a brand name, that's how I differentiate my products--matrix multiplication products--from the competition. Because I'm fast/faster/fastest.
I always interpreted this work as a satire of Ayn Randian style objectivism/exceptionalism. It's funny that in school we were taught that it was a criticism of communism/socialism when I think it's really a criticism of the popular perception of communism/socialism.
The underlying premise is a misunderstanding of equity that persists today. The idea behind leftist ideologies is not to handicap the brilliant and able, but to extend the material buffers that allow brilliance to flourish to everyone:
“I am, somehow, less interested in the weight and convolutions of Einstein's brain than in the near certainty that people of equal talent have lived and died in cotton fields and sweatshops.”
But at the time and today, people (leftists included) thought that leftist ideology must have been based in envy or some inferiority complex. Harrison Bergeron is that 'meta-ideology' taken to its extreme.
Yep. I am a junior engineer. I got a message on Linkedin, someone was offering me like $2k/week to take interviews/technical assessments for clients overseas. I asked if it was legal and he stopped messaging me.
I always wondered how their initial python web scraper was fast enough to index the entire internet on that old hardware, even given the small size of the internet at the time. I guess the answer is that they had a local backup at Stanford. Thanks for sharing!
Politics and war aside, I think it's cool that the profile of the B-2/B-21 airframe closely resembles a falcon in a dive. The control scheme of the falcon is similar too. I wonder how much of it is a coincidence and how much is a sort of convergent evolution.
All an agent has to do is query their current spot in the graph and it will return a vector that leads them to the next lowest cost. This is useful if you have lots of agents going to the same location.
https://www.youtube.com/watch?v=BHcQ4JCj27w
The description of this video has a lot of good resources. I made it when I was a much much worse programmer though so I wouldn't bother actually watching the video lol.