Livoa LogoLivoa
b
Input dataset
x₁
x₂
y
1.2
6.1
3.0
33.3
2.2
?
2D TabPFN layer (12x)
1D feature attention
1D sample attention
MLP
Predictions: ŷ_test
a
TabPFN is trained on synthetic data to take entire


datasets as inputs and predict in a forward pass

Xtrain
TabPFN


neural network

parameterized by θ

-log qθ(ytest | ...)


Training loss to be optimized

across millions of datasets

b
2D TabPFN layer (12x)
1D feature attention
1D sample attention
Xtrain
TabPFN


neural network

parameterized by θ

ytest
33.3
6.1
2D TabPFN layer (12x)
1D feature attention
1D sample attention
MLP

2

by faf

0
0 uses