Why do very deep plain CNNs (without skip connections) sometimes show higher training error than shallower CNNs?
They always overfit because parameter count increases.
Optimization becomes harder; skip connections make learning identity mappings easier and improve gradient flow.
BatchNorm guarantees this never happens.
Using sigmoid activations instead of ReLU always fixes it.
MobileNetV1 reduces computation primarily by using:
Dilated convolutions everywhere
Depthwise separable convolutions (depthwise + pointwise)
ROI pooling
Fully-connected layers instead of convolutions
EfficientNet’s key scaling idea is:
Only scale depth while keeping width and resolution fixed
Only scale resolution while keeping depth and width fixed
Compound scaling of depth, width, and resolution using a single coefficient
Scale width but reduce depth to keep parameters constant
Sign in with Google — it is free — to see every question with its answer and explanation, practise it in learning mode, or take it as a timed mock test.
The IIT Madras BS Deep Learning for Computer Vision (Deep Learning for Computer Vision) Quiz 2 paper sat on 16 Aug 2026, in the May 2026 term: 22 questions for 50 marks in 120 minutes. The first 3 questions are below. Sign in with Google — it is free — to see the whole paper with its answers and explanations, in learning mode or as a timed mock test.
| Feature | Deep Learning for Computer Vision Quiz 2 16 Aug 2026 at a glance |
|---|---|
| Term | May 2026 term |
| Subject | Deep Learning for Computer Vision |
| Course code | BSDA5006 |
| Questions | 22 |
| Marks | 50 |
| Duration | 120 min |
| MCQ | 13 |
| MSQ | 3 |
| Numerical | 6 |
| Official paper | Deep Learning For Computer Vision 16 Aug 26 |
| Negative marking | No negative marking. |
| Updated |
