>We have successfully replaced thousands of complicated deep net time series based anomaly detectors at a FANG with statistical (nonparametric, semiparametric) process control ones.
They use 3 to 4 orders lower number of trained parameters and have just enough complexity that a team of 3 or four can handle several thousands of such streams.
Could you explain how ? Cause I am working on this essentially right now and it seems management is wanting to go the way of Deep NNs for our customers.
Without knowing details it's very hard to give specific recommendations. However if you follow that thread you will see folks have commented on what has worked for them.
In general I would recommend get Hyndman's (free) book on forecasting. That will definitely get you upto speed.
If it's the case that you will ship the code over client's fence and be done with it, that is, no commitments regarding maintenance, then I will say do what the management wants. If you will continue to remain responsible for the ongoing performance of the tool then you will be better if choosing a model you understand.
MBAs do love their neural nets. As a data scientist you have to figure out what game you’re playing: is it the accuracy game or the marketing game? Back when I was a data scientist, I got far better results from “traditional” models than NN, and I was able to run off dozens of models some weeks to get a lot of exposure across the org. Combined with defensible accuracy, this was a winning combination for me. Sometimes you just have to give people what they want, and sometimes that’s cool modeling and a big compute spend rather than good results.
Without getting into specifics (just joined a new firm), we’re working with a bunch of billing data.
Management is leaning toward a deep learning forecasting approach — train a neural net to predict expected cost and then use multiple deviation scorers (including Wasserstein distance) to flag anomalies.
A simpler v1 is already live, and this newer approach isn’t my call. I’m still fairly new to anomaly detection, so for now I’m mostly trying to learn and ship within the existing direction rather than fight it.
There is no single answer, because there are multiple architectures for foundation time-series models, such as T5, decoder-only models, and state-space models (SSMs).
For Chronos-2 (the current state of the art in time-series modeling), the setup is almost identical to that of LLMs because it is based on the T5 architecture. The main difference is that, in time-series models, tokens correspond to subintervals in the real-valued (ℝ) space. You can check the details here: https://arxiv.org/pdf/2510.15821
Then they will break and wear off quite fast I imagine.
Take a look at industrial cobots (not a typo). They feature rounded corners, have very little to no "finger pinchy areas" and lots of force feedback sensors.
Despite that they still require trained (adult) personal and move very slowly when actually interacting with humans.
That's the price for them being sturdy and precise.
I only knew him from directing Harry Met Sally and Wolf of Wall street where almost all of his scenes are hillarious, especially the one where he burst into the room abusing DiCaprio and his gang over expenses.
He directed The Princess Bride, This Is Spinal Tap, When Harry Met Sally, A Few Good Men, and Misery. Didn't know this, but he directed a sequel to This Is Spinal Tap.
Yes,I know BUT of his personal works, those two remain the only ones I have seen.
And also A Few Good Men (I did not know it was one of his works till now)
I mean this had to be obvious right... I just wonder of what the current TAM is, and how much can Open AI crack that.
Cause the problem is Gemini and Claude are not the default for anyone except us tech geeks and even then I default to ChatGPT most of the time, but I use Claude for coding.
Are they confident they can meet the revenue projections they have made like 200B by 2027 simply through this?
That are 4 words. ;-) Other then that yes, goverment contracts, meaning offloading the costs to the regular people. I think the US goes all in with AI, trying to get world dominant and controlling the market. But I think this is a mistake and the US will get into trouble with this.
They use 3 to 4 orders lower number of trained parameters and have just enough complexity that a team of 3 or four can handle several thousands of such streams.
Could you explain how ? Cause I am working on this essentially right now and it seems management is wanting to go the way of Deep NNs for our customers.
reply