Glossary¶

A short reference for the spatial-coding vocabulary used throughout neurospatial. Each entry says what the term means, names the neurospatial concept it maps to, and links to the closest notebook.

Allocentric¶

A reference frame fixed in the world, not the animal. Allocentric coordinates are the cartesian (x, y) or (x, y, z) of physical space; an allocentric bearing is measured against compass directions (east is the standard zero in neurospatial). Contrast with egocentric. See ops.egocentric.allocentric_to_egocentric and 24_object_vector_cells.

Bayesian decoding¶

Reading the animal's position back out from a population of neurons. Given a vector of spike counts across cells in a short window, Bayesian decoding computes a probability ("posterior") over every spatial bin: how likely the animal is to be there given the observed firing pattern. From the posterior you can extract a single best-guess location (the MAP estimate, the bin with the highest probability) or the posterior mean. The discrepancy between the estimate and the animal's actual position is the decoding error. The required inputs are a rate map per cell (the "encoding model" — what each cell fires across space, built from a training segment of the recording) and the spike counts during the segment you want to decode. See decode_position and 20_bayesian_decoding.

Bin¶

A single discrete cell of the discretized environment. Each bin carries a fixed area (2D) or volume (3D), a centroid in the allocentric frame, and a node ID in the connectivity graph. neurospatial identifies bins by integer index throughout; see Environment.bin_at. Bins are sometimes called nodes when the graph view is more relevant — e.g. when computing geodesic distances.

Cell (graph)¶

Synonym for bin in graph contexts. neurospatial uses "cell" only when interfacing with libraries that prefer the graph vocabulary; the canonical term in the public API is bin.

Egocentric¶

A reference frame fixed in the animal's body. The egocentric origin is the animal's current position, and angles are measured relative to its current heading: 0 = ahead, π/2 = left, -π/2 = right. Used to describe how the animal perceives nearby objects regardless of which compass direction it's currently facing. See compute_egocentric_rate, 22_spatial_view_cells, and 24_object_vector_cells.

Field¶

A scalar quantity defined per bin: shape (n_bins,) for single-time quantities, (n_time, n_bins) for time-varying. Place fields, boundary fields, occupancy maps, and decoded posteriors are all fields. See compute_spatial_rate which returns a rate field.

Linearization¶

Mapping a 2D position onto a 1D coordinate by projecting it onto a track graph (a linear graph embedded in 2D, like a T-maze or W-track). The 1D coordinate respects topological adjacency on the track even when bins are far apart in 2D space. See Environment.to_linear and 05_track_linearization.

Object-vector cell (OVC)¶

A neuron that fires whenever a specific object is at a specific distance and egocentric direction from the animal. The tuning curve is a 2D firing-rate map indexed by (distance, egocentric bearing) — the object-vector. See compute_egocentric_rate, is_object_vector_cell, and 24_object_vector_cells.

Occupancy¶

The amount of time the animal spent in each bin: shape (n_bins,) with units of seconds. Bins the animal never visited have occupancy 0 (or NaN, depending on the consumer). Firing-rate computation divides spike counts by occupancy bin-wise.

Place cell¶

A neuron whose firing rate is selective for the animal's location in the environment, irrespective of head direction or behavior. The spatial firing-rate map of a place cell typically has one or a few compact place fields. See compute_spatial_rate, detect_place_fields, and 11_place_field_analysis.

Place field¶

A contiguous region of high firing rate in a place cell's rate map. Detected via thresholding + connected-component analysis; quantified by location, size, peak rate, and information content. See detect_place_fields.

Rate map¶

The spatial firing-rate field for one neuron: shape (n_bins,) in units of Hz. Built from spike counts ÷ occupancy, optionally smoothed via graph diffusion or Gaussian KDE. The phrase "place field" refers to a thresholded subset of the rate map.

Spike-triggered¶

Anything indexed by spike times rather than time bins. A spike-triggered average of position gives the animal's expected location at the moment each spike fired; a spike-triggered event-aligned histogram (PSTH) does the analogous slicing around event times. See peri_event_histogram and 26_peri_event_psth.

Tuning curve¶

A 1-D firing-rate function over some behavioral variable other than position — head direction, speed, time-from-event. Computed by binning the variable, counting spikes per bin, and dividing by occupancy. Direction tuning is the most common case. See compute_directional_rate and 25_head_direction_tuning.

View field¶

A spatial firing-rate field indexed by the location the animal is looking at, not where the animal is standing. Built from gaze-direction trajectories and a view model (fixed-distance, ray-cast, or boundary-intersection). See compute_view_rate, is_spatial_view_cell, and 22_spatial_view_cells.