The specific (fictitious) example they gave was $100 -> $150 due to a ~$23 tariff.
> But if we bump the cost of freight, insurance, and customs from $5 to, say, $28, then they wholesale the shoes to Footlocker for about $75. And if Footlocker purchases Nike shoes for $75, then they retail them for $150. Everyone needs to fixed percentages to avoid losses.
The point being that many parts of the supply chain don't operate on fixed costs and instead percentages.
> The point being that many parts of the supply chain don't operate on fixed costs and instead percentages.
That's how it was working in most wealthy countries, where tariffs were generally low or very low.
I don't see why it must continue to work that way in the USA with 50% or 100% or more tariffs. If Footlocker wish to charge double the post-tariff price, that leaves room for a competitor to change double the pre-tariff price.
(Or double the pre-tariff plus a tiny bit, to account for the increased cost of insurance, theft etc.)
1. Only for the first version, not for this version. I am sorry!
2. Yeah ours is guaranteed ok, as we wrote code to generate it basically just from plain torch ops. The code to run inference is available, just not the training code and data generation.
3. We have put it to work on time series data, which is very business relevant for example https://github.com/liam-sbhoo/tabpfn-time-series, and we have a table in the Appendix with all datasets we evaluate on in our main analysis to give you some ideas for possible datasets.
“Yeah ours is guaranteed ok, as we wrote code to generate it basically just from plain torch ops.”
This is where there might be claims. It already sounds safer than training on copyrighted works. The only thing that could remain is if it was a derivative work by reusing parts of copyrighted works in your process.
So, I’m curious about how you produced the specifications that the data was generated from. In my case, I was going to just use open versions of all kinds of equations that I’d hand-convert to internal representations. Others might be fair use if my description were high level enough that it wasn’t close to theirs. Some I couldn’t use at all because they were patented and independent versions are prohibited by law.
Did you all also derive your causal models from real-world formulas and data sets? If so, did you have a rule about putting distance between your representation and theirs? Or was it an entirely-random, search process across endless configurations? (I have a hard time imagining the latter would work.)
Yes, there are normalizations applied before the features are fed to the neural network. Additionally, the neural network is trained on a very diverse set of artificial datasets.
Author here: The new introduction of attention between features did make a big impact compared to the first variant of TabPFN. The old model handled every feature like it was completely different to be feature 5 vs 15, but actually features are typically more-or-less permutation invariant. So the logic is similar to why a CNN is better for images than an MLP.
To put it very simply, the trick is that while the others train a new model for each problem, TabPFN is pre-trained to handle any kind of problem on the fly.
To draw a parallel to NLP: previously people trained a neural network for each kind of text classification they wanted to do, but then LLMs came around that pre-trained to learn to perform new tasks on the fly.
Similarly, TabPFN learns to do new tasks on the fly just from the context (dataset) given.
Training and prediction in these models is by default one and the same, similar to how the prediction of the next token in an LLM is not split into learning from context and then doing the actual prediction.
There is a way to split this even up, though, then the predictions, I believe, take something like 1/10s for medium-sized datasets.
No, it is *much* stronger, a different architecture and scales to 10x the number of examples. It can also do regression now, and handle categorical features. Please, have a quick look at the abstract before making such claims.
There's a lazy way to submit to arxiv, which is to submit just the PDF, even if you did it in latex. Sometimes it can be annoying to organize the tex files to submit to arxiv. It's uncommon, but the font and math rendering are the standard latex font.
