W

Definition: A distance function used in statistics and machine learning to measure the distance between two probability distributions. It is particularly useful in scenarios where you want to compare distributions in a more intuitive way than traditional methods like the KL-divergence.

Definition: Also known as narrow AI, it refers to AI systems that are designed to handle one particular task or set of tasks. Weak AI is the opposite of strong AI, which is designed to perform any intellectual task that a human can.

Definition: A regularisation technique used in neural networks and machine learning to prevent overfitting. Weight decay involves adding a penalty to the loss function equivalent to the sum of the weights squared, effectively reducing the magnitude of the weights in the model.

Definition: Parameters within a neural network that are adjusted during training. Weights determine the influence that one node in a neural network has on another node in the next layer.

Definition: The process of setting the initial values of the weights in a neural network before training begins. Proper initialisation can help speed up the convergence of training and lead to better overall performance.

Definition: A preprocessing technique used to transform input data so that its covariance matrix is the identity matrix, meaning that features become less correlated and all have the same variance. This is often done before applying certain machine learning algorithms to make the data easier to work with.

Definition: Refers to a type of neural network architecture that connects all or most of the inputs directly to the output layer, in contrast to deep learning, which uses multiple hidden layers between inputs and outputs.

Definition: In signal processing, a mathematical function used to select a subset of a signal. In machine learning, window functions are used to extract features from sequences of data.

Definition: A type of word representation that allows words with similar meaning to have a similar representation. It is a learned representation for text where words that have the same meaning have a similar representation.

Definition: A group of related models used to produce word embeddings. These models are shallow, two-layer neural networks that are trained to reconstruct linguistic contexts of words. Word2Vec takes as its input a large corpus of text and produces a vector space, with each unique word in the corpus being assigned a corresponding vector in the space.

Definition: The design, execution, and automation of processes based on workflow rules where human tasks, data, or files are routed between people or systems based on pre-defined business rules.

Definition: A feature selection technique that uses a predictive model to score feature subsets according to their predictive power. The method “wraps” this model evaluation within the feature selection process.

Definition: A caching technique in which updates to the cache cause corresponding updates to the underlying storage. This technique is used to ensure that the cache always contains the most up-to-date information and is commonly used in databases and AI applications where data integrity is critical.

Definition: A process in neural network training where the smallest weights (those closest to zero) are systematically removed from the model. This is a form of model simplification that can improve efficiency and reduce the size of the model without significantly reducing accuracy.

Definition: A collection of machine learning algorithms for data mining tasks. It contains tools for data pre-processing, classification, regression, clustering, association rules, and visualisation, and is freely available for educational and research purposes.