Stolen probability: A structural weakness of neural language models

Keyphrases

• Structural Weakness 100%
• Neural Language Models 100%
• Embedding Space 75%
• Distance Metric 50%
• Dot Product 50%
• Inductive Bias 50%
• Probability Distribution 25%
• Convex Hull 25%
• Softmax 25%
• Product Distance 25%
• Word Vector 25%
• Prediction Vector 25%
• Optimal Order 25%
• High-dimensional Embedding 25%

Mathematics

• Probability Theory 100%
• Neural Network 100%
• Dot Product 66%
• Probability Distribution 33%
• Assigning Probability 33%
• Part Form 33%
• Softmax Function 33%
• Convex Hull 33%

Psychology

• Neural Network 100%
• Probability Distribution 33%