Module 4 Assignment: CNN architecture memo

Module 4 Assignment: CNN architecture memo#

Theme#

Convolutional neural networks for vision

Scenario#

A computer-vision team is deciding whether a small CNN is adequate for grayscale inspection images before moving to larger pretrained models.

Exercises#

  1. Design a CNN for a small grayscale image classification problem.

  2. Specify convolution, activation, pooling, flattening, and output-head choices.

  3. Run the starter model and inspect output shapes after each stage.

  4. Explain how local structure and weight sharing support the design.

Evidence Requirements#

  • A short explanation of the data, tensors, objective, and evaluation signal used in the starter experiment.

  • At least one meaningful modification to the starter code, with the changed variable named explicitly.

  • A comparison against the unmodified starter result or another defensible baseline.

  • A limitation statement that separates what the toy experiment demonstrates from what a production model would require.

Submission#

Submit a 600-900 word technical memo plus code, plots, tables, or shape traces needed to support your claims. The memo should read like a review artifact for another AI practitioner: concise, reproducible, and honest about uncertainty.

Rubric Focus#

  • Technical correctness and appropriate neural-network vocabulary.

  • Evidence from the starter experiment or a documented extension.

  • Connection between design choices and data/problem structure.

  • Clear treatment of limitations, failure modes, or next experimental gates.

import torch
from torch import nn

model = nn.Sequential(
    nn.Conv2d(1, 4, kernel_size=3, padding=1), nn.ReLU(), nn.MaxPool2d(2),
    nn.Conv2d(4, 8, kernel_size=3, padding=1), nn.ReLU(), nn.AdaptiveAvgPool2d((1, 1)),
    nn.Flatten(), nn.Linear(8, 2)
)

x = torch.randn(5, 1, 16, 16)
for layer in model:
    x = layer(x)
    print(f"{layer.__class__.__name__:>18}: {tuple(x.shape)}")

            Conv2d: (5, 4, 16, 16)
              ReLU: (5, 4, 16, 16)
         MaxPool2d: (5, 4, 8, 8)
            Conv2d: (5, 8, 8, 8)
              ReLU: (5, 8, 8, 8)
 AdaptiveAvgPool2d: (5, 8, 1, 1)
           Flatten: (5, 8)
            Linear: (5, 2)

Reflection Prompts#

  • What changed when you modified the starter experiment, and why should that change matter?

  • Which result surprised you, and what diagnostic would you run next?

  • What assumption would you document before handing this model to another practitioner?

  • Which failure mode from the module reading is most relevant to your result?