Unlocking Tokens as Data Points for Generalization Bounds on Larger Language Models.
Sanae LotfiYilun KuangBrandon AmosMicah GoldblumMarc FinziAndrew Gordon WilsonPublished in: CoRR (2024)
Keyphrases
- language model
- generalization bounds
- data points
- data dependent
- learning theory
- generalization ability
- language modeling
- linear classifiers
- model selection
- vc dimension
- ranking algorithm
- document retrieval
- probabilistic model
- query expansion
- learning problems
- test collection
- information retrieval
- nearest neighbor
- hyperplane
- euclidean space
- statistical learning theory
- high dimensional
- dimensionality reduction
- feature space
- low dimensional
- learning tasks
- euclidean distance
- high dimensional data
- support vector machine
- data distribution
- learning machines
- bp neural network
- ranking functions
- em algorithm
- kernel machines
- labeled data