histc_op

→TensorImplPtr

int histc_op(const int & a, int bins, double lo, double hi)

source edit

Python wrappers

pylucid.ops.histcfunction

Histogram counts over $[\mathrm{lo}, \mathrm{hi})$ with bins uniform buckets.

Composite — counts-only wrapper around histogram_op. When lo == hi the range is auto-derived from the input via min_op and max_op, with the degenerate min == max case bumped by $+1$ so the bins remain well-defined (matches reference framework behaviour). Not differentiable — integer-valued output.

Math

Let $w = (\mathrm{hi} - \mathrm{lo}) / \mathrm{bins}$ . Then for each bin index $k \in [0, \mathrm{bins})$ ,

y_k = \#\!\left\{\, i : \mathrm{lo} + k w \le a_i < \mathrm{lo} + (k+1) w \,\right\}

See Also

histogram_op — full version that also returns bin edges.

Parameters

aTensorImplPtr

Input tensor of arbitrary shape.

binsint64

Number of histogram bins. Must be positive.

lodouble

Inclusive lower edge. If lo == hi, auto-range from min(a) to max(a).

hidouble

Exclusive upper edge.

Returns

TensorImplPtr

1-D tensor of length bins holding integer counts.

Raises

Failure

If a is null.

Notes

Values outside [lo, hi) are dropped silently by the underlying histogram_op — no overflow / underflow bins are returned.