Definition:
Parameter /pəˈræ.mɪ.tə/ noun — In machine learning and statistical modeling, a parameter is a model-internal variable that is learned from the training data and determines how the model processes inputs to make predictions or classifications.
Parameters are central to the learning process. During training, algorithms adjust these values to minimize a loss function, thereby improving model performance. In neural networks, for instance, parameters include:
- Weights — values assigned to connections between neurons
- Biases — constants added to neuron activations
Unlike hyperparameters, which are set before training begins, parameters are automatically tuned by the model during training using optimization techniques like gradient descent.
Well-tuned parameters enable a model to:
- Capture meaningful patterns
- Generalize effectively to unseen data
- Respond dynamically to different input scenarios
Parameters are foundational to any predictive or generative machine learning model.
« Back to dictionary
