TransferBenchmark<T>

Models & Types · AiDotNet.SelfSupervisedLearning.Evaluation

Transfer learning benchmark for evaluating SSL representations on downstream tasks.

For Beginners

Transfer learning benchmarks test how well pretrained representations transfer to new tasks. We take an encoder pretrained on one dataset (e.g., ImageNet) and evaluate it on different tasks (e.g., object detection, segmentation).

How It Works

Common transfer benchmarks:

Evaluation protocols:

Constructors

Constructor Summary
TransferBenchmark(INeuralNetwork<T> encoder, int encoderOutputDim) Initializes a new instance of the TransferBenchmark class.

Methods

Method Summary
BenchmarkResult<T> KNNEvaluation(Tensor<T> trainData, int[] trainLabels, Tensor<T> testData, int[] testLabels, int k) Runs k-NN evaluation on a downstream dataset.
BenchmarkResult<T> LinearProbing(Tensor<T> trainData, int[] trainLabels, Tensor<T> testData, int[] testLabels, int numClasses, int epochs) Runs linear probing evaluation on a downstream dataset.
BenchmarkSuite<T> RunFullSuite(Tensor<T> trainData, int[] trainLabels, Tensor<T> testData, int[] testLabels, int numClasses, string datasetName) Runs a full benchmark suite.
Dictionary<double, BenchmarkResult<T>> FewShotEvaluation(Tensor<T> trainData, int[] trainLabels, Tensor<T> testData, int[] testLabels, int numClasses, double[] percentages) Runs few-shot evaluation with limited labeled data.