TabPFNNetwork<T>
Models & Types · AiDotNet.NeuralNetworks.Tabular
TabPFN (Prior-Fitted Network) neural network for tabular data.
Inheritance: NeuralNetworkBase<T> → TabularNeuralNetworkBase<T> → TabPFNNetwork<T>
For Beginners
TabPFN is unique because it learns from synthetic data:
Architecture:
- Input Projection: Convert features to embeddings
- Deep Transformer: Many attention layers (typically 12)
- In-Context Learning: Training data provided as context
- Output Prediction: Classification directly from attention
Key insight: TabPFN was trained on millions of synthetic classification tasks, learning to “learn” from examples. At inference, you provide training data as context and it predicts test labels without fine-tuning.
This makes TabPFN extremely fast for small datasets since there’s no traditional training loop - just a forward pass with context.
How It Works
TabPFN is a transformer model trained on synthetic data that can perform in-context learning on tabular classification tasks. It approximates Bayesian inference through attention mechanisms. This implementation follows the standard neural network architecture pattern with:
- Proper inheritance from NeuralNetworkBase
- Layer-based architecture using ILayer components
- Engine-based tensor operations for CPU/GPU acceleration
- Full autodiff and JIT compilation support
Reference: “TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second” (2022)
Example
var architecture = new NeuralNetworkArchitecture<float>(inputFeatures: 8, outputSize: 4);
var options = new TabPFNOptions<double> { NumFeatures = 20, MaxContextSamples = 1000 };
var input = Tensor<float>.CreateRandom(new[] { 1, 20 });
var trainX = Tensor<float>.CreateRandom(4, 8);
var trainY = Tensor<float>.CreateRandom(4, 2);
var result = new AiModelBuilder<float, Tensor<float>, Tensor<float>>()
.ConfigureModel(new TabPFNNetwork<float>(architecture))
.Build(trainX, trainY);
var output = result.Predict(input);
Constructors
| Constructor | Summary |
|---|---|
TabPFNNetwork() |
Initializes a new instance with default architecture settings. |
TabPFNNetwork(NeuralNetworkArchitecture<T> architecture, TabPFNOptions<T> options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction) |
Initializes a new TabPFN network with the specified architecture. |
Properties
| Property | Summary |
|---|---|
TabPFNOptions<T> Options { get; } |
Gets the TabPFN-specific options. |
int EmbeddingDimension { get; } |
Gets the embedding dimension. |
int NumLayers { get; } |
Gets the number of transformer layers. |
Methods
| Method | Summary |
|---|---|
Dictionary<string, T> GetFeatureImportance() |
|
ModelMetadata<T> GetModelMetadata() |