applicable models Filters

Browse and search through available filters in this category. Click on any filter to view related techniques.

Description
Action
black box
Only requires input-output access, no internal model access needed
76
model agnostic
Works with any model type without requiring specific architecture (black-box techniques)
73
supervised
Requires labelled training data
50
white box
Requires full model transparency and access
35
neural networks
Techniques for general neural network architectures
26
training data
Requires access to the original training dataset
24
gradient access
Requires access to model gradients
18
llm
Techniques for Large Language Models (GPT, BERT, etc.)
16
parametric
Models with a fixed number of parameters
16
discriminative
Models that learn decision boundaries directly
12
generative
Models that learn data distributions
12
model internals
Requires access to weights, neurons, or internal representations
12
probabilistic output
Model must provide probability distributions as output
10
differentiable
Model must be differentiable
9
transformer
Techniques for transformer-based architectures
6
linear models
Techniques for linear and generalized linear models
5
unsupervised
Works with unlabeled data
5
architecture specific
Requires specific architectural components to function
5
tree based
Techniques for tree-based algorithms (decision trees, random forests, gradient boosting)
3
convolutional
Techniques for CNNs and vision models
2
Showing 20 of 34 filters