Algorithm documentation

Fits a straight line that minimizes squared prediction errors.

Maps a weighted feature score to a class probability.

Splits the feature space into rule-based regions.

Combines diverse decision trees trained on sampled data.

Predicts from the labels or values of nearby observations.

Finds a separating boundary with the widest possible margin.

Partitions observations around movable cluster centroids.

Rotates data onto directions that preserve the most variance.

Adds weak models sequentially to correct earlier residual errors.

Learns layered nonlinear transformations through connected units.

Predicts class probabilities using feature independence assumption.

Groups points that are closely packed together into dense regions.

Builds a hierarchy of clusters using distance metrics.

Fits a regression model while shrinking large coefficients to reduce overfitting.

Represents data as overlapping probability clouds with soft cluster membership.

Finds anomalies by measuring how quickly random splits isolate each observation.

Compresses an input through a bottleneck and learns to reconstruct it.

Learns the long-term value of actions by exploring rewards in an environment.

Computes how much every network weight contributed to an error, then updates it.

Places similar high-dimensional observations near one another on a 2D map.

Builds a compact vocabulary by repeatedly merging frequent neighboring symbols.

Maps discrete token IDs to dense vectors whose geometry can encode meaning.

Adds a position-dependent signal so identical tokens at different places remain distinguishable.

Lets each token gather information from other tokens according to learned relevance.

Prevents each token from looking at future tokens during autoregressive generation.

Runs several attention projections in parallel so different relationships can be learned.

Alternates contextual attention with token-wise nonlinear processing and residual paths.

Reuses earlier key and value vectors instead of recomputing the whole prompt every step.

Turns model logits into a probability distribution and chooses the next token.

Keeps several high-scoring partial sequences instead of committing to one token at a time.

Repeatedly swaps adjacent values that are in the wrong order.

Selects the smallest remaining value and places it into its final position.

Builds a sorted prefix by inserting each new value into its correct place.

Recursively sorts halves, then merges them in order.

Partitions values around a pivot, then recursively sorts both sides.

Builds a max heap and repeatedly moves its root to the sorted suffix.

Runs insertion sort across shrinking gaps before the final adjacent pass.

Counts each integer value, then emits values in numeric order.

Stably groups integers by one digit at a time, from least to most significant.

Distributes values into ranges, sorts each bucket, and concatenates them.