Learn Before
Probability Distribution Output of an Encoder-Classifier Model
The output of a classification model, mathematically represented as for an input , is a probability distribution over a predefined set of labels (e.g., ). This conditional distribution is denoted by . To determine the final classification, the model selects the label associated with the highest probability in this calculated distribution.
0
1
Tags
Foundations of Large Language Models
Ch.1 Pre-training - Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Related
A machine learning model is designed to classify movie reviews as 'positive' or 'negative'. The model uses a two-part structure: an initial component transforms the raw text of a review into a numerical summary, and a second component takes this summary and assigns the final 'positive' or 'negative' label. The model performs well on reviews it was trained on, but when given new reviews with slightly different vocabulary (e.g., using 'brilliant' instead of 'excellent'), it classifies them incorrectly, even though the numerical summaries it generates for these new reviews are very similar to the summaries of positive reviews it has seen before. Which of the following is the most likely explanation for this issue?
A system for identifying fraudulent financial transactions operates in a two-stage process. First, it transforms raw transaction data into a meaningful summary of behavior patterns. Second, it uses this summary to make a final judgment. Arrange the following events into the correct logical order that represents this process.
Diagnosing a Text Classification Model
Probability Distribution Output of an Encoder-Classifier Model