trainingObjective
P12747
predicate
Indicates the goal or target outcome that a training process is designed to achieve.
Observed surface forms (4)
- lossFunction ×5
- optimizationObjective ×4
- trainingGoal ×2
- learningObjective ×1
Sample triples (22)
| Subject | Object |
|---|---|
| A3C | maximize expected cumulative reward via predicate surface "optimizationObjective" ⓘ |
| AlexNet | cross-entropy loss via predicate surface "lossFunction" ⓘ |
| AlphaZero | maximize expected game outcome via predicate surface "learningObjective" ⓘ |
| Boltzmann machines | maximize data log-likelihood ⓘ |
| CLIP | InfoNCE-style loss via predicate surface "lossFunction" ⓘ |
| CLIP | contrastive loss via predicate surface "lossFunction" ⓘ |
| CLIP | maximize similarity of matching image-text pairs ⓘ |
| CLIP | minimize similarity of non-matching image-text pairs ⓘ |
| GPT-2 | next token prediction ⓘ |
| GPT-3 | next-token prediction ⓘ |
| GPT-3.5 | next-token prediction ⓘ |
| Generative Adversarial Networks | adversarial loss via predicate surface "lossFunction" ⓘ |
| LeNet | classification accuracy via predicate surface "optimizationObjective" ⓘ |
| LeNet | cross-entropy loss via predicate surface "lossFunction" ⓘ |
| LogisticRegression | logistic loss minimization with regularization via predicate surface "optimizationObjective" ⓘ |
| MuZero | maximize expected cumulative reward via predicate surface "optimizationObjective" ⓘ |
| Oregon State Beavers men’s golf | development of collegiate golfers ⓘ |
| WaveNet | cross-entropy loss over quantized samples ⓘ |
| WaveNet | maximum likelihood estimation ⓘ |
|
WebText dataset
surface form:
WebText
|
next-token prediction ⓘ |
| lua (Hawaiian martial art) | battlefield effectiveness via predicate surface "trainingGoal" ⓘ |
| lua (Hawaiian martial art) | rapid incapacitation of opponents via predicate surface "trainingGoal" ⓘ |