andy
3fc7fa863e
Affected files: .obsidian/graph.json .obsidian/workspace-mobile.json .obsidian/workspace.json Christmas Movies.md Politcs/Fascism.md Politcs/Neoliberalism/Neoliberalism.md STEM/AI/Neural Networks/CNN/Interpretation.md STEM/CS/Languages/Python.md STEM/IOT/OS/Composition.md STEM/IOT/OS/Contiki.md STEM/IOT/OS/OS.md STEM/IOT/OS/README.md STEM/img/iot-event-based.png STEM/img/iot-os-stack.png STEM/img/iot-thread-based.png
1.3 KiB
1.3 KiB
Activation Maximisation
- Synthesise an ideal image for a class
- Maximise 1-hot output
- Maximise SoftMax
- Use trained network
- Don't update weights
- Feedforward noise
- Back-propagate loss
- Don't update weights
- Update image
- Back-propagate loss
Regulariser
- Fit to natural image statistics
- Prone to high frequency noise
- Minimise
- Total variation
x^*
is the best solution to minimise loss
x^*=\text{argmin}_{x\in \mathbb R^{H\times W\times C}}\mathcal l(\phi(x),\phi_0)
- Won't work
x^*=\text{argmin}_{x\in \mathbb R^{H\times W\times C}}\mathcal l(\phi(x),\phi_0)+\lambda\mathcal R(x)
- Need a regulariser like above
\mathcal R_{V^\beta}(f)=\int_\Omega\left(\left(\frac{\partial f}{\partial u}(u,v)\right)^2+\left(\frac{\partial f}{\partial v}(u,v)\right)^2\right)^{\frac \beta 2}du\space dv
\mathcal R_{V^\beta}(x)=\sum_{i,j}\left(\left(x_{i,j+1}-x_{ij}\right)^2+\left(x_{i+1,j}-x_{ij}\right)^2\right)^{\frac \beta 2}
- Beta
- Degree of smoothing