linear_probe.hpp File Reference

Linear probe – trains one linear classifier to decode a binary concept from a layer's activations, answering "is this concept linearly represented here?". More...

#include <cmath>
#include <cstdint>
#include <random>
#include <stdexcept>
#include <string>
#include <vector>
#include "pulsatrix/determinism.hpp"
#include "pulsatrix/bce_with_logits_loss.hpp"
#include "pulsatrix/linear_module.hpp"

Include dependency graph for linear_probe.hpp:

Go to the source code of this file.

Classes
class	pulsatrix::LinearProbe
	A linear probe: LinearModule(activation_dim, 1) + BCEWithLogitsLoss, trained on (activation, binary concept label) pairs. High post-training accuracy means the concept is linearly decodable from those activations; chance-level accuracy means it is not (at least not linearly). More...

Namespaces
namespace	pulsatrix

Detailed Description

Linear probe – trains one linear classifier to decode a binary concept from a layer's activations, answering "is this concept linearly represented here?".