PyTorch中Tensor的查找和筛选

本文源码基于版本1.0，交互界面基于0.4.1

import torch

按照指定轴上的坐标进行过滤

index_select()
沿着某tensor的一个轴dim筛选若干个坐标

>>> x = torch.randn(3, 4)		# 目标矩阵
>>> x
tensor([[ 0.1427,  0.0231, -0.5414, -1.0009],
        [-0.4664,  0.2647, -0.1228, -1.1068],
        [-1.1734, -0.6571,  0.7230, -0.6004]])
>>> indices = torch.tensor([0, 2])	# 在轴上筛选坐标
>>> torch.index_select(x, dim=0, indices)	# 指定筛选对象、轴、筛选坐标
tensor([[ 0.1427,  0.0231, -0.5414, -1.0009],
        [-1.1734, -0.6571,  0.7230, -0.6004]])
>>> torch.index_select(x, dim=1, indices)
tensor([[ 0.1427, -0.5414],
        [-0.4664, -0.1228],
        [-1.1734,  0.7230]])

where()
用于将两个broadcastable的tensor组合成新的tensor，类似于c++中的三元操作符“?:”

>>> x = torch.randn(3, 2)
>>> y = torch.ones(3, 2)
>>> torch.where(x > 0, x, y)
tensor([[1.4013, 1.0000],
        [1.0000, 0.9267],
        [1.0000, 0.4302]])
>>> x
tensor([[ 1.4013, -0.9960],
        [-0.3715,  0.9267],
        [-0.7163,  0.4302]])

指定条件返回01-tensor

>>> x = torch.arange(5)   
>>> x
tensor([0, 1, 2, 3, 4])
>>> torch.gt(x,1)		# 大于
tensor([0, 0, 1, 1, 1], dtype=torch.uint8)
>>> x>1					# 大于
tensor([0, 0, 1, 1, 1], dtype=torch.uint8)
>>> torch.ne(x,1)		# 不等于
tensor([1, 0, 1, 1, 1], dtype=torch.uint8)
>>> x!=1				# 不等于
tensor([1, 0, 1, 1, 1], dtype=torch.uint8)
>>> torch.lt(x,3)		# 小于
tensor([1, 1, 1, 0, 0], dtype=torch.uint8)
>>> x<3					# 小于
tensor([1, 1, 1, 0, 0], dtype=torch.uint8)
>>> torch.eq(x,3)		# 等于
tensor([0, 0, 0, 1, 0], dtype=torch.uint8)
>>> x==3				# 等于
tensor([0, 0, 0, 1, 0], dtype=torch.uint8)

返回索引

>>> x = torch.arange(5)
>>> x   # 1维
tensor([0, 1, 2, 3, 4])
>>> torch.nonzero(x)
tensor([[1],
        [2],
        [3],
        [4]])
>>> x = torch.Tensor([[0.6, 0.0, 0.0, 0.0],[0.0, 0.4, 0.0, 0.0],[0.0, 0.0, 1.2, 0.0],[0.0, 0.0, 0.0,-0.4]])
>>> x   # 2维
tensor([[ 0.6000,  0.0000,  0.0000,  0.0000],
        [ 0.0000,  0.4000,  0.0000,  0.0000],
        [ 0.0000,  0.0000,  1.2000,  0.0000],
        [ 0.0000,  0.0000,  0.0000, -0.4000]])
>>> torch.nonzero(x)
tensor([[0, 0],
        [1, 1],
        [2, 2],
        [3, 3]])

借助nonzero()我们可以返回符合某一条件的index（https://stackoverflow.com/questions/47863001/how-pytorch-tensor-get-the-index-of-specific-value）

>>> x=torch.arange(12).view(3,4)
>>> x
tensor([[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]])
>>> (x>4).nonzero()
tensor([[1, 1],
        [1, 2],
        [1, 3],
        [2, 0],
        [2, 1],
        [2, 2],
        [2, 3]])