Classify a new row based on the decision tree T.

Free the buffer without destroying the elements of the sequence. This is intended for use after the contents of the sequence are relocated

Find such lambda that if used for regularization, optimum would have all weights zero.

The conditional information of a (with labels) based on b (a continuous, although discrete variable). It picks the cut of b that minimizes this info

info of a conditioned on b

these are bucket sizes used by the default allocator. They are powers of two starting at 16 and going until SystemMemory / 64. Note that them being p

Calculates the direct forces between all pairs of particles in two near-field leaf nodes. Directly updates forces in Left, and places forces for ngh

Calculates the direct interactions between and within leaves. Since the forces are symmetric, this calculates the force on one side (rightNeighbors) w

Translate from exterior (multipole) expansion to interior (local) expansion along all far-field interactions.

Translate and accumulate interior (local) expansions down the tree, including applying them to all particles at the leaves.

Determine if two nodes (boxes) are separated enough to use the multipole approximation.

Given an a ray, a bounding box, and a sequence of triangles, returns the index of the first triangle the ray intersects inside the box. The triangles

calculate the force at a nearby point based on expansion

Calculates forces, placing them in particles[i].force

n is the number of events (i.e. twice the number of triangles)

these two only work on internal nodes

puts the leaves of tree tr into the array Leaves in left to right order

Returns the number of blocks in the flattened range

Pop from local stack.

A "staged" version of in_cicle. It takes 3 points, returning a function on one more point that reports whether that point is in the circumcircle of th

information content of s (i.e. entropy * size)

Determine the number of workers to spawn

Implement initialize_default manually rather than calling initialize_fill(n, value_type()) because this allows us to store a sequence of uncopyable ty

Pre calculate feature (column) covariants and A^T * y.

prioritized linear probing a new key will bump an existing key up if it has a higher priority an equal key will replace an old key if replaceQ(new,old

The following two functions are the core of the CK method. They calculate the "well separated decomposition" of the points.

Returns true if the sequence is in small size mode

The main algorithm

Generate a random graph of size n