New Posts  All Forums:

Posts by xlink

coding...It's pretty easy to eat up RAM if you're doing machine learning or AI.
If they're using anything SORT of like the original neural network algorithm, shouldn't they be able to use imprecise computation to set up some sort of proprotype network and then use that as an initialization point for traditional forward/backward propagation?
The people who started the testing that way probably left the company 20 years ago.What you have is a mostly new set of guys who just kept on doing what those before did without thinking about it."Here's the procedure for doing it in the US, report these everywhere"When in doubt assume incompetence before malfeasance, especially when dealing with anything REMOTELY concerned with a government.
I don't have an MBA(though I'll admit that I plan on getting one) and my background is in STEM.I am biased because I do work at a company with a heavy tech focus.
It averages out...http://mathaholic.com/images/lawoflargenumbers.jpgAlso there IS such a thing as stochastic load balancing where you balance your selections away from the heavily loaded system..
MBAs don't like lawsuits.Again... counter examples to what you're stating are Amazon, Google, Facebook, Netflix and a few others. They all have their share of MBAs.Stating that MBAs are the problem requires A LOT of very strong assumptions which in their totality are invalid.
Money is a major constraint. It isn't the ONLY constraint.In the short run, labor is fixed... this matters when you have deadlines to meet (esp. if you're dealing with external parties)Same applies with capital. Hard to run simulations or build prototypes when your hardware is maxed out.The hope would be that all resources are allocated optimally given the constraints at play. Obviously there are a lot of unknowns and impossible to measure obstacles.
MBAs are still the problem?Yeah dude... you've obviously never worked with a team of engineers, many of whom have great ideas about hemorrhaging edge technologies and who want to incorporate many of them into a production environment.It isn't MBAs per se, it's a combination of non-technical people making technical decisions and overly technical people getting too tied up in their own world to realize that they're harming overall system health.If it were MBAs, Amazon(heavy...
My experience is that layoffs mostly affect middle management right now.I've seen a lot of decent people go over the last few years... mostly Sr. Managers and Directors. Handful of people at the bottom whose jobs were getting automated but mostly middle management.The ONE guy was really good whose job got eliminated ended up getting converted to a principle engineer (same pay). He's competent and touching a revenue affecting system.
You realize that timings refer to the number of cycles which actions are delayed by...so 4333 Mhz with a latency of 20 cycles is similar to 2166@ 10which is similar to 1083 @ 5which is similar to 542@ 2.5which is similar to 433 @ 2If you told someone that you got a set of winbond BH5 running @ 433 with 2 CL you'd have A LOT of people who are impressed.Now I understand that this isn't really IMPROVING latency... but when your CPU has something like 20mb cache... you won't...
New Posts  All Forums: