Indexing and slicing tensors in TensorFlow - Model Pipeline Trace

Model Pipeline - Indexing and slicing tensors

This pipeline shows how a tensor (multi-dimensional array) is accessed and sliced to extract smaller parts. This is like cutting a cake into pieces to share.

Data Flow - 5 Stages

1Input tensor

1 tensor of shape (4, 5)→Create a 2D tensor with 4 rows and 5 columns→1 tensor of shape (4, 5)

[[1, 2, 3, 4, 5], [6, 7, 8, 9, 10], [11, 12, 13, 14, 15], [16, 17, 18, 19, 20]]

↓

2Indexing a single element

tensor of shape (4, 5)→Access element at row 2, column 3 (0-based indexing)→scalar (single number)

Element at [2, 3] is 14

↓

3Slicing rows

tensor of shape (4, 5)→Select rows 1 to 3 (excluding 3), all columns→tensor of shape (2, 5)

[[6, 7, 8, 9, 10], [11, 12, 13, 14, 15]]

↓

4Slicing columns

tensor of shape (4, 5)→Select all rows, columns 2 to 4 (excluding 4)→tensor of shape (4, 2)

[[3, 4], [8, 9], [13, 14], [18, 19]]

↓

5Slicing with step

tensor of shape (4, 5)→Select every other row, all columns→tensor of shape (2, 5)

[[1, 2, 3, 4, 5], [11, 12, 13, 14, 15]]

Training Trace - Epoch by Epoch

No training loss to show for tensor slicing operations

Epoch	Loss ↓	Accuracy ↑	Observation
1	N/A	N/A	No training, only tensor slicing operations

Prediction Trace - 5 Layers

Layer 1: Input tensor

Layer 2: Indexing element [2,3]

Layer 3: Slicing rows 1:3

Layer 4: Slicing columns 2:4

Layer 5: Slicing rows with step 0:4:2

Model Quiz - 3 Questions

What is the shape of the tensor after slicing rows 1 to 3 (excluding 3) from a (4, 5) tensor?

A (3, 5)

B (1, 5)

C (2, 5)

D (4, 3)

Key Insight

Indexing and slicing tensors lets you pick parts of data easily, like cutting a cake into slices. This is essential for preparing data for machine learning models.