return to table of content

TimesFM: Time Series Foundation Model for time-series forecasting

whimsicalism
34 replies
1d2h

I'm curious why we seem convinced that this is a task that is possible or something worthy of investigation.

I've worked on language models since 2018, even then it was obvious why language was a useful and transferable task. I do not at all feel the same way about general univariate time series that could have any underlying process.

wuj
5 replies
22h49m

Time series data are inherently context sensitive, unlike natural languages which follow predictable grammar patterns. The patterns in time series data vary based on context. For example, flight data often show seasonal trends, while electric signals depend on the type of sensor used. There's also data that appear random, like stock data, though firms like Rentech manage to consistently find unlerlying alphas. Training a multivariate time series data would be challenging, but I don't see why not for specific applications.

Xcelerate
4 replies
5h2m

Is Rentech the only group that genuinely manages to predict stock price? Seems like the very observation that it’s still possible would be enough motivation for other groups to catch up over such a long period.

Also, the first realistic approximation of Solomonoff induction we achieve is going to be interesting because it will destroy the stock market.

icapybara
1 replies
3h52m

Agreed, if stock prices were predictable by some technical means, they would be quickly driven to unpredictability by people trading on those technical indicators.

frankc
0 replies
2h23m

This is that old finance chestnut. Two finance professors are walking down the hall and one of them spots a twenty dollar bill. He goes to pick it up but the other professor stops him and says "no don't bother. If there was twenty dollars there someone would have already picked it up"

Yes, people arbitrage away these anomalies, and make billions doing it.

amelius
0 replies
58m

Maybe that would be a good thing. I wouldn't mourn the destruction of the stock market as it's just a giant wealth-gap increasing casino. Trading has nothing to do with underlying value.

zeroxfe
3 replies
22h55m

I'm curious why we seem convinced that this is a task that is possible or something worthy of investigation.

There's a huge industry around time series forecasting used for all kinds of things like engineering, finance, climate science, etc. and many of the modern ones incorporate some kind of machine learning because they deal with very high dimensional data. Given the very surprising success of LLMs in non-language fields, it seems reasonable that people would work on this.

whimsicalism
2 replies
21h5m

Task specific time series models, not time series “foundation models” - we are discussing different things.

zeroxfe
0 replies
16h48m

I don't think we are. The premise of this is that the foundation model can learn some kind of baseline ability to reason about forecasting, that is generalizable across different domains (each which needs fine tuning.) I don't know if it will find anything, but LLMs totally surprised us, and this kind of thing seems totally worthy of investigation.

cscurmudgeon
0 replies
14h11m

Foundational time series models have been around since 2019 and show competitive levels of performance with task specific models.

https://arxiv.org/abs/1905.10437

shaism
3 replies
22h8m

Fundamentally, the pre-trained model would need to learn a "world model" to predict well in distinct domains. This should be possible not regarding compute requirements and the exact architecture.

After all, the physical world (down to the subatomic level) is governed by physical laws. Ilya Sutskever from OpenAI stated that next-token prediction might be enough to learn a world model (see [1]). That would imply that a model learns a "world model" indirectly, which is even more unrealistic than learning the world model directly through pre-training on time-series data.

[1] https://www.youtube.com/watch?v=YEUclZdj_Sc

whimsicalism
1 replies
21h3m

But the data generating process could be literally anything. We are not constrained by physics in any real sense if we predicting financial markets or occurrences of a certain build error or termite behavior.

shaism
0 replies
20h38m

Sure, there are limits. Not everything is predictable, not even physics. But that is also not the point of such a model. The goal is to forecast across a broad range of use cases that do have underlying laws. Similar to LLM, they could also be fine-tuned.

wavemode
0 replies
4h22m

"predicting the next token well means that you understand the underlying reality that led to the creation of that token"

People on the AI-hype side of things tend to believe this, but I really fundamentally don't.

It's become a philosophical debate at this point (what does it mean to "understand" something, etc.)

itronitron
3 replies
19h49m

There was a paper written a while back that proved mathematically how you can correlate any time series with any other time series, thus vaporizing any perception of value gained by correlating time series (at least for those people that read the paper.) just wanted to share

notnaut
0 replies
19h8m

I would like to read more. Feels sort of like an expression of certain “universal truths” like the 80/20 rule or golden ratio

bdjsiqoocwk
0 replies
19h11m

What does that mean "you can correlate"? That phrase is meaningless.

bigger_cheese
3 replies
12h39m

There is potential for integrating ML with time series data in industrial applications (things like smelters, reactors etc.), where you have continuous stream of time series measurements from things like gauges and thermocouples. If you can detect (and respond) to changing circumstances faster then a humans in control room reacting to trends or alarms then potential big efficiency gains...

Operator guidance is often based on heuristics - when metric A exceeds X value for Y seconds take action Z. Or rates of change if the signal is changing at a rate of more than x etc.

So in these areas there exists potential for ML solution, especially if it's capable of learning (i.e. last response overshot by X so trim next response appropriately).

kqr
2 replies
10h1m

Every time i've actually tried something like this it has not outperformed statistical process control.

It's not just that control charts are great signal detectors, but also managing processes like that takes a certain statistical literacy one gets from applying SPC faithfully for a while, and does not get from tossing ML onto it and crossing fingers.

chaos_emergent
1 replies
3h49m

Every time i've actually tried something like this it has not outperformed statistical process control.

There are clear counterexamples to your experience, most notably in maintaining plasma stability in tokamak reactors: https://www.nature.com/articles/s41586-021-04301-9

whimsicalism
0 replies
3h30m

task specific model

smokel
1 replies
22h56m

The things that we are typically interested in have very clear patterns. In a way, if we find that there are no patterns, we don't even try to do any forecasting.

"The Unreasonable Effectiveness of Mathematics in the Natural Sciences" [1] hints that there might be some value here.

[1] https://en.m.wikipedia.org/wiki/The_Unreasonable_Effectivene...

yonixw
0 replies
22h16m

Exactly, so for example, I think the use of this model is in cases where you want user count to have some pattern around timing. And be alerted if it has spike.

But you wouldn't want this model for file upload storage usage which only increases, where you would put alerts based on max values and not patterns/periodic values.

refibrillator
1 replies
15h16m

Why do you think language is so special?

There's an extensive body of literature across numerous domains that demonstrates the benefits of Multi-Task Learning (MTL). Actually I have a whole folder of research papers on this topic, here's one of the earliest references on hand that I feel captures the idea succinctly in the context of modern ML:

“MTL improves generalization by leveraging the domain-specific information contained in the training signals of related tasks" [Caruana, 1998]

I see repetition and structure everywhere in life. To me it's not far fetched that a model trained on daily or yearly trends could leverage that information in the context of e.g. biological signals which are influenced by circadian rhythm etc.

Disclaimer: my background is in ML & bio-signals, I work with time series too much.

owl_brawl
0 replies
14h24m

For those who haven't read it, Rich Caruana's thesis on multi-task learning is beautifully written (the cited 1998 paper here). It's amazing to see how far the field has come, and, at the same time, how advanced the thinking was in the 90s too.

cma
1 replies
6h53m

Watch this talk from Albert Gu:

Efficiently Modeling Long Sequences with Structured State Spaces

https://www.youtube.com/watch?v=luCBXCErkCs

They made one of the best time series models and it later became one of the best language models too (Mamba).

whimsicalism
0 replies
3h28m

I have already watched that talk and know Albert Gu. His work is not about a “foundational” time series model but rather a task specific one.

sarusso
0 replies
1d1h

+1 for “any underlying process”. It would be interesting what use case they had in mind.

matt-p
0 replies
8h46m

Not really. It's true it would usually need more context than a single series dataset but you can predict broadly accurate-ish bandwidth usage trends just using simple statistical extrapolation, we've been doing that since the early 90s. If you give a model your subscriber numbers and usage data as time series it should be able to tell you quite accurately how much electricity|bandwidth|gas|road traffic levels| metro passenger levels at station Z... you'll be using at 4pm on January 4th 2026.

klysm
0 replies
5h44m

I think there are some generalizable notions of multiscale periodicity that could get embedded into some kind of latent space.

fedeb95
0 replies
11h21m

as you say, without knowing anything about the underlying process, we can't predict generally. Some other comments point to contexts in which we do know something about the underlying. For instance, I don't think finance is something where you can apply this kind of stuff.

baq
0 replies
23h44m

well... if you look at a language in a certain way, it is just a way to put bits in a certain order. if you forget about the 'language' part, it kinda makes sense to try because why shouldn't it work?

IshKebab
0 replies
23h38m

Why not? There are plenty of time series that have underlying patterns which means you can do better than a total guess even without any knowledge of what you are predicting.

Think about something like traffic patterns. You probably won't predict higher traffic on game days, but predicting rush hour is going to be pretty trivial.

nwoli
12 replies
1d3h

Seems like a pretty small (low latency) model. Would be interesting to hook up to mouse input (x and y) and see how well it predicts where I’m gonna move the mouse (maybe with and without seeing the predicted path)

throwtappedmac
10 replies
1d2h

Curious George here: why are you trying to predict where the mouse is going? :)

tasty_freeze
4 replies
1d1h

Game developers are constantly trying to minimize lag. I have no idea if computers are so fast these days that it is a "solved" problem, but I knew a game developer ages ago who used a predictive mouse model to reduce the apparent lag by guessing where the mouse would be at the time the frame was displayed (considering it took 30 ms or whatever to render the frame).

ukuina
0 replies
17h5m

What an amazing look into one of the greatest minds in programming!

Thank you for this treasure.

The relevant bits:

I am now allowing the client to guess at the results of the users movement until the authoritative response from the server comes through. This is a biiiig architectural change. The client now needs to know about solidity of objects, friction, gravity, etc. I am sad to see the elegent client-as-terminal setup go away, but I am practical above idealistic.

The server is still the final word, so the client is allways repredicting it's movement based off of the last known good message from the server.
wongarsu
0 replies
1d1h

Competitive online games commonly predict the player's movement. Network latencies have improved and are now usually <16ms (useful milestone since at 60fps you render a frame every 16.6ms), but players expect to still be able to smoothly play when joining from the other side of the continent to play with their friends. You usually want every client to agree where everyone is, and predicting movement leads to less disagreement than what you would get from using "outdated" state because of speed-of-light delays.

If you want to predict not just position but also orientation in a shooter game, that's basically predicting the mouse movements.

orbital-decay
0 replies
1d1h

The only thing worse than lag is uneven lag, which is what you're going to end up with. Constant lag can be dealt with by players, jitter can't.

teaearlgraycold
1 replies
1d2h

Think of the sweet sweet ad revenue!

throwtappedmac
0 replies
1d2h

Haha as if advertisers don't know me better than I know me

nwoli
1 replies
1d2h

Just to see how good the model is (maybe it’s creepily good in a fun way)

Timon3
0 replies
1d1h

There's a fun game idea in there! Imagine having to outmaneuver a constantly learning model. Not to mention the possibilities of using this in genres like bullet hell...

brigadier132
0 replies
22h49m

Catching cheaters in games might seem like a good use.

jarmitage
0 replies
1d3h

What is the latency?

esafak
8 replies
1d1h

Is anyone using neural networks for anomaly detection in observability? If so, which model and how many metrics are you supporting per core?

sarusso
3 replies
1d1h

What do you mean by “observability”?

tiagod
0 replies
21h26m

Depending on how stable your signal is, I've had good experience with seasonal ARIMA and LOESS (but it's not neural networks)

sarusso
0 replies
1d1h

Oh, yes I am working on that. Usually LSTM, exploring encoder-decoders and generative models, but also some simpler models based on periodic averages (which are surprisingly useful in some use cases). But I don’t have per-core metrics.

morkalork
2 replies
1d

How data hungry is it, or what is the minimum volume of data needed before its worth investigating?

viraptor
1 replies
15h49m

The more complex the data is, the more you need. If your values are always 5, then you need only one data point.

morkalork
0 replies
4h6m

If your values were always 5,you wouldn't use an LSTM to model it either. So presumably there's a threshold for when LSTM becomes practical and useful, no?

polskibus
6 replies
1d1h

how good is it on stocks?

svaha1728
3 replies
1d

The next index fund should use AI. What could possibly go wrong?

whimsicalism
2 replies
1d

I promise you your market-making counterparties already are.

hackerlight
1 replies
19h58m

What kind of things are they doing with AI?

whimsicalism
0 replies
19h36m

Predicting price movements, finding good hedges, etc.

fedeb95
0 replies
11h19m

it doesn't apply. Checkout the Incerto by Nassim Nicholas Taleb.

claytonjy
0 replies
21h46m

if I knew it was good, why would I tell you that?

dangerclose
6 replies
1d3h

is it better than prophet from meta?

VHRanger
4 replies
1d2h

I imagine they're both worse than good old exponential smoothing or SARIMAX.

Pseudocrat
3 replies
1d2h

Depends on use case. Hybrid approaches have been dominating the M-Competitions, but there are generally small percentage differences in variance of statistical models vs machine learning models.

And exponentially higher cost for ML models.

VHRanger
1 replies
1d1h

At the end of the day, if training or doing inference on the ML model is massively more costly in time or compute, you'll iterate much less with it.

I also think it's a dead end to try to have foundation models for "time series" - it's a class of data! Like when people tried to have foundation models for any general graph type.

You could make foundation models for data within that type - eg. meteorological time series, or social network graphs. But for the abstract class type it seems like a dead end.

rockinghigh
0 replies
1d

These models may be helpful if they speed up convergence when fine tuned on business-specific time series.

SpaceManNabs
0 replies
23h50m

is there a ranking of the methods that actually work on benchmark datasets? Hybrid, "ML" or old stats? I remember eamonnkeogh doing this on r/ML a few years ago.

efrank3
0 replies
15h28m

Prophet was pretty bad so yes, but it doesn't seem much better than ARIMA

claytonjy
2 replies
22h25m

And like all deep learning forecasting models thus far, it makes for a nice paper but is not worth anyone using for a real problem. Much slower than the classical methods it fails to beat.

p1esk
1 replies
21h29m

That’s what people said about CV models in 2011.

claytonjy
0 replies
21h4m

That's fair, but they stopped saying it about CV models in 2012. We've been saying this about foundational forecasting models since...2019 at least, probably earlier. But it is a harder problem!

toasted-subs
0 replies
23h18m

Something I've had issues with time series has been having to use relatively custom models.

It's difficult to use off the shelf tools when starting with math models.

iamgopal
4 replies
1d3h

How can time series model be pre-trained ? I think I’m missing something.

melenaboija
1 replies
1d3h

Third paragraph of the introduction of the mentioned paper[1] in the first paragraph of the repo.

[1] https://arxiv.org/abs/2310.10688

jurgenaut23
0 replies
1d2h

I guess they pre-trained the model to exploit common patterns found in any time-series (e.g., seasonalities, trends, etc.)... What would be interesting, though, is to see if it spots patterns that are domain-specific (e.g., the ventricular systole dip in an electrocardiogram), and possibly transfer those (that would be obviously useless in this specific example, but maybe there are interesting domain transfers out there)

sarusso
0 replies
1d1h

My understating is that, while your eye can naturally spot a dependency over time in time series data, machines can’t. So as we did for imaging, where we pre-trained models to let machines easily identify objects in pictures, now we are doing the same to let machines “see” dependencies over time. Then, how these dependencies work, this is another story.

malux85
0 replies
1d1h

If you have a univariate series, just single values following each other -

[5, 3, 3, 2, 2, 2, 1, …]

What is the next number? Well let’s start with the search space - what is the possible range of the next number? Assuming unsigned 32bit integers (for explanation simplicity) it’s 0-(2^32-1)

So are all of those possible outputs equally likely? The next number could be 1, or it could be 345,654,543 … are those outputs equally likely?

Even though we know nothing about this sequence, most time series don’t make enormous random jumps, so no, they are not equally likely, 1 is the more likely of the two we discussed.

Ok, so some patterns are more likely than others, let’s analyse lots and lots of time series data and see if we can build a generalised model that can be fine tuned or used as a feature extractor.

Many time series datasets have repeating patterns, momentum, symmetries, all of these can be learned. Is it perfect? No, but what model is? And things don’t have to be perfect to be useful.

There you go - that’s a pre-trained time series model in a nutshell

uoaei
3 replies
1d4h

"Time series" is such an over-subscribed term. What sorts of time series is this actually useful for?

For instance, will it be able to predict dynamics for a machine with thousands of sensors?

sarusso
1 replies
1d1h

Even if it was for multivariate time series, the model would first need to infer what machine are we talking about, then its working conditions, and only then make a reasonable forecast based on an hypothesis of its dynamics. I don’t know, seems pretty hard.

uoaei
0 replies
23h57m

Indeed. An issue I ran into over and over while doing research for semiconductor manufacturing.

My complaint was more illustrative than earnest.

techwizrd
0 replies
1d3h

Specifically, its referring to univariate, contiguous point forecasts. Honestly, I'm a little puzzled by the benchmarks.

optimalsolver
2 replies
1d1h

When it comes to time series forecasting, if the method actually works, it sure as hell isn't being publicly released.

speedgoose
0 replies
22h29m

Some times series are more predictable than others. Being good at predicting the predictable ones is useful.

For example you can easily predict the weather with descent accuracy. Tomorrow is going to be about the same than today. From there you can work on better models.

Or predicting a failure in a factory because a vibration pattern on an industrial machine always ended up in a massive failure after a few days.

But I agree that if a model is good at predicting the stock market, it’s not going to be released.

baq
0 replies
23h43m

and yet we have those huge llamas publicly available. these are computers that talk, dammit

aantix
2 replies
23h22m

Would this be useful in predicting lat/long coordinates along a path? To mitigate issues with GPS drift.

If not, what would be a useful model?

bbstats
0 replies
16h53m

Kalman filter

DeathArrow
2 replies
9h55m

It seems to me that predicting something based on time is rarely accurate and meaningful.

Suppose you want to buy stocks? Would you look on a time based graph and buy according to that? Or you rather look at financial data, see earnings, profits? Wouldn't a graph that has financial performance on x-axis be more meaningful that one that has time?

What if you research real estate in a particular area? Wouldn't be square footage a better measure than time?

Terretta
1 replies
8h27m

Would you look on a time based graph and buy according to that? Or you rather look at financial data, see earnings, profits?

Things affecting financials happen through time.

DeathArrow
0 replies
5h22m

All things happen through time, but my argument is that time might not be the best parameter to model relations.

chaos_emergent
1 replies
21h33m

"Why would you even try to predict the weather if you know it's going to be wrong?"

- most OCs on this thread

david_shi
0 replies
33m

I have a few qualms with this app: 1. For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem. From Windows or Mac, this FTP account could be accessed through built-in software.

2. It doesn't actually replace a USB drive. Most people I know e-mail files to themselves or host them somewhere online to be able to perform presentations, but they still carry a USB drive in case there are connectivity problems. This does not solve the connectivity issue.

3. It does not seem very "viral" or income-generating. I know this is premature at this point, but without charging users for the service, is it reasonable to expect to make money off of this?

celltalk
1 replies
12h36m

If I give this model the first 100 prime numbers, does it give me back the rest of it? If so what is the circuit?

fedeb95
0 replies
11h20m

how is the series of the first 100 prime numbers a time series ?

viraptor
0 replies
15h53m

I'm not sure I understand two things here. Could someone clarify: 1. This is a foundation model, so you're expected to fine tune for your use case, right? (But readme doesn't mention tuning) 2. When submitting two series, do they impact each other in predictions?

mhh__
0 replies
1d

Dear googler or meta-er or timeseries transformer startup something-er: Please make a ChatGPT/chat.lmsys.org style interface for one of these that I can throw data at and see what happens.

This one looks pretty easy to setup, in fairness, but some other models I've looked at have been surprisingly fiddly / locked behind an API.

Perhaps such a thing already exists somewhere?

htrp
0 replies
5h25m

Prophet 2.0

hm-nah
0 replies
13h47m

Anyone have insights working with Ikigai’s “Large Graphical Model” and how well it does on time-series? It’s proprietary, but I’m curious how well it performs.