Link to this week’s scratchpad: 01_Week1_Jan_260115_123946.pdf
This week I spent most of my time reviewing the concepts from my Machine learning class because I wanted to present a machine learning paper in my research group’s journal club.
-
The paper was @laydiNovelAttentionMechanism2025 which present a method to remove background noise from images of Microtubules obtained in live cell imaging.
-
I reviewed basic concepts like back propagation and somewhat advanced models like U-Net.
-
One thing I realized was that I didn’t practice the concepts presented during my machine learning course by hand: that is use the Feynmann technique to explain to myself, for example, how each iteration during the training step looks like for an architecture like ResNet, for example.
-
Learned about squeeze and excitation networks and how they work by assigning an ‘importance coefficient’ to each channel of the data during training.
-
Learned that SciPy uses Levenberg-Marquardt algorithm for curve-fitting. need to learn more about this.
1. Friction & Gaps
- What blocked me: Started learning a lot of topics at the same time didn’t finish any of them completely lol.
- What I’ll try next week:
2. Signals
- Recurring themes:
- Promote to permanent note: Legendre Transforms, Squeeze and Excitation