The Einsum notation makes it desirable to formulate your model/layer as multi-dimensional arrays connected by (loosely) named axes, without worrying too much about breaking it down to primitives yourself. Once you get used to it, the terseness is liberating.
Some implementations of TC/ESC will abruptly brake one or more of the wheels in on/off pulses, using the ABS pump, disrupting an otherwise balanced cornering car.
Yes. I won't buy a car I cannot disable ABS on, because it extends braking distances, plus by doing so, all traction control is off.
But I honed my non-ABS driving in my teenage years, pre-ABS (for most cars at the time), on rural dirt roads, and on roads with constant snow and ice, and driving on frozen lakes.
Of course, if you disable ABS, you may end up without any form of real differential, as many cars use ediff tech, which is horrible, so I now have to also vet for a true hardware, non-open diff.
ABS has nothing to do with improving braking distance. Any skilled driver, who knows their car, can do far better.
For skilled drivers, ABS extends braking distance.
What ABS does, is let a driver mash the pedal on full, and still steer around obstacles. Something a skilled driver can still do.
One problem with these tests, is that if you disable ABS on some cars, the proportioning valve is still set to 50/50, front/rear, meaning you start to lose traction on the rear wheels, and skid. This causes a loss of control, and reduced braking power.
Cars prior to ABS has the capability to adjust rear brake pressure by weight, a typical default of 70/30.
Some card have ABS controllers which you can disable ABS functionality, but still retain proportioning control. Cars with proper diffs are often like this.
Is this true with a modern abs, and have you tested your own capabilities vs abs? It can act on individual wheels which no human can do with a regular brake pedal.
I know of ABS capabilities, and extensively tested ABS on/off on every car I own.
ABS is about steering when people slam the brake pedal down, it is not about improving braking distance.
Outside of how it doesn't help on pavement, it is a absolute disaster on gravel, and deep snow.
For example, on gravel if you lock up the brakes, you dig in. Gravel builds in front of the tire, and your tire sinks. ABS won't allow this, and so on gravel I can stop from high speed fast, while ABS actively works to deny my ability to stop.
On snow, if you briefly lock up the brakes, snow builds in front of the wheel. You can then spin the wheels to turn, let up, and the car will instantly take off in a new direction. ABS actively prevents this.
ABS was never, ever designed to reduce braking distance. It was designed to allow people to steer while braking.
The video I linked shows drivers (one professional, one not) achieved stopping distance in a straight line with and without abs in a fairly recent car.
Gravel and snow I understand - but at least for me are pretty big exceptions.
I did the same, hoping that someday I will finally get FTTH.
Currently still on the only real option here which is Comcast 1G/20M. Sometimes LTE has better upload speed (but not latency, obviously).
If you're referring to column mounted prndl stalks, they don't turn with the wheel. Examples: Mercedes, older Model S, and lots of pickup trucks.
If you're referring to paddle shifters, ironically Ferrari (of wheel-mounted indicator buttons fame) mounts them on the column instead (again, not turning with the wheel). IMHO both are fine for street and track usage because manual shifting is most likely triggered when the car is (almost) straight. However this assumption does not apply to turning indicator controls, which by definition are used when you are, uh, turning or about to turn!
Precisely, Tesla went their own way for no reason other than cost.. it is objectively worse than their previous implementation, the mainstream implementation, and any of the competing less-standard implementations that Tesla defenders are grasping at.
There's a reason you don't put control buttons on moving parts.
It's like putting the elevator buttons on the door.
Want to press your floor, please wait until the door shuts.. or lunge at it as the door moves. Want to hit the door close button.. sorry that's no longer possible.
Imagine motorized blinds where the button is on the blinds. Better get the ladder out when they are up & you want to bring them down.
Imagine a fan where the buttons were on the spinning blades, or the center rotor with buttons moved while on.
I don't understand the design decisions. You know he was heavily involved in some Tesla change when its something this dumb because no one can stop him.
Those column-mounted paddle shifters also span a large radius on both sides, which combined with their quick steering ratio makes them effectively available at all times aside from when you're putting in a lot of steering lock – which is the one scenario you shouldn't be shifting aggressively anyways.
Silly thought: so it's shoehorning evolution into organizations. Regardless of how something is made to replicate itself, it inherently does so because that's what it does. Evolution arises when changes (accidental or not, from some perspective) present opportunities for thriving. It's not a matter of will as it is a matter of fact. I suppose for organizations, regulations and oversight are necessary to prevent evolution towards fulfilling perverse incentives. A bit hard to do when we're dealing with hyacinths, though. Herbicide, anyone?
Pournelle was just another Republican and this is just another political slogan, not an "iron law".
He was also an engineer, which makes it even worse, because it means he has old engineer brain where you decide you know everything about everyone else's fields.
Get any undersink RO system. I went with a tanked system (there is a pressurized tank for storing filtered water) but I have friends who were happy with tankless (tap gives you the trickle of filtered water directly).
I live in an area with very hard tap water (500 ppm TDS); even without a water softener the output becomes ~80 ppm TDS, and the water hardness test strips check out at ~40 ppm. I then installed a water softener with its own sediment pre-filter, and now the overall RO output looks (TDS-wise) even more pure than your typical plastic-bottled water.
I refuse to do takehomes as any part of the interview. My principle is out of fairness and mutual respect of the value of time: For every hour I spend interviewing, someone from the company must spend 1 hour too.
EDIT: Take-home-styled coding exercise in a pair programming setup is fine --- I've taken a few of those and they were actually quite pleasant. The task is not the issue --- respect is.
