mindspore.ops.nextafter

mindspore.ops.nextafter(input, other)[source]

Returns the next representable floating-point value after input towards other element-wise.

Say there are two float32 numbers \(a\), \(b\), and let the representable delta of float32 datatype is \(eps\). If \(a < b\), then the next representable of \(a\) towards \(b\) is \(a+eps\), the next representable of \(b\) towards \(a\) is \(b-eps\).

\[out_{i} = nextafter({input_{i}, other_{i}})\]

For more detailed information, refer to A Self Regularized Non-Monotonic Neural Activation Function.

Parameters

  • input (Tensor) – The first input tensor. The shape of tensor is \((N,*)\) where \(*\) means, any number of additional dimensions. Must be one of the following types: float32, float64.

  • other (Tensor) – The second input tensor. The shape of tensor is \((N,*)\) where \(*\) means, any number of additional dimensions. Must be one of the following types: float32, float64.

Returns

Tensor, has the same shape and data type as input.

Raises

  • TypeError – If neither input nor other is a Tensor.

  • TypeError – If the dtype of input and other is not one of: float32, float64.

  • TypeError – If the dtypes of input and other are not same.

  • ValueError – If input’s shape is not the same as other.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore
>>> import numpy as np
>>> from mindspore import Tensor, ops
>>> input_ = Tensor(np.asarray([0.0]), mindspore.float32)
>>> other_ = Tensor(np.asarray([0.1]), mindspore.float32)
>>> output_ = ops.nextafter(input_, other_)
>>> print(output_)
[1.e-45]