# Lecture - Theories of Deep Learning MT25, XI, Visualising the filters and response in a CNN

> Source: https://ollybritton.com/notes/uni/part-c/mt25/theories-of-deep-learning/lectures/cnn/ · Updated: 2025-11-19 · Tags: uni, lecture

- [Course - Theories of Deep Learning MT25](https://ollybritton.com/notes/uni/part-c/mt25/theories-of-deep-learning/)

- Transfer learning: Download a pre-existing net and then re-train the last layer, or last few layers

![Screenshot 2025-11-18 at 13.44.06.png](https://ollybritton.com/assets/attachments/img/Screenshot 2025-11-18 at 13.44.06.png)

- Spectra of the net through the learning process
- Why does transfer learning work?
- When does transfer learning not work?
	- Consider e.g. fine-tuning a network that learns to classify dogs vs cats in order to classify specific breeds: this is not a good idea since the model has explicitly learned to map out differences in dog and cat breeds

![Screenshot 2025-11-19 at 11.28.41.png](https://ollybritton.com/assets/attachments/img/Screenshot 2025-11-19 at 11.28.41.png)

- CNNs learn representations resembling wavelets / curvelets / ridgelets

- Attention sinks
-

---
Olly Britton — https://ollybritton.com. Machine-readable index: https://ollybritton.com/llms.txt