},
"active": "3e78d839de43b5f0",
"lastOpenFiles": [
- "Pasted image 20251106010304.png",
- "Pasted image 20251106010018.png",
- "Pasted image 20251106010007.png",
- "Pasted image 20251106005956.png",
- "Pasted image 20251106005919.png",
- "Pasted image 20251106005851.png",
- "Pasted image 20251106005803.png",
- "Pasted image 20251106005742.png",
- "Pasted image 20251106005719.png",
- "Pasted image 20251106005655.png",
- "Pasted image 20251106005628.png",
+ "Pasted image 20251106230015.png",
+ "Pasted image 20251106225651.png",
+ "Pasted image 20251106225645.png",
+ "Pasted image 20251106225012.png",
+ "Pasted image 20251106224648.png",
+ "Pasted image 20251106224445.png",
+ "Pasted image 20251106224411.png",
+ "Pasted image 20251106224329.png",
+ "Pasted image 20251106223635.png",
+ "Pasted image 20251106222907.png",
+ "Pasted image 20251106222245.png",
"University/Machine Learning/Full Notes.md",
"Untitled 1.md",
"some_ideas.md",
http://peterbloem.nl/blog/pca
-(there is something about Convolutional neural networks but it seems to be optional - I think I went to this last time actually but hey)
\ No newline at end of file
+(there is something about Convolutional neural networks but it seems to be optional - I think I went to this last time actually but hey)
+
+## Convolutional Neural Networks
+introduction to deep learning
+
+feature extraction vs deep leearning
+![[Pasted image 20251106130821.png]]
+
+end-to-end learning: end goal (output) used to learn feature extraction (input)
+
+![[Pasted image 20251106130852.png]]
+
+![[Pasted image 20251106130914.png]]
+
+![[Pasted image 20251106130935.png]]
+![[Pasted image 20251106130951.png]]
+
+
+![[Pasted image 20251106130959.png]]
+
+![[Pasted image 20251106131009.png]]
+
+![[Pasted image 20251106131033.png]]
+
+![[Pasted image 20251106131119.png]]
+![[Pasted image 20251106131137.png]]
+
+![[Pasted image 20251106131147.png]]
+
+
+![[Pasted image 20251106131201.png]]
+
+![[Pasted image 20251106131249.png]]
+non-linearity in this case things like ReLU or such
+
+![[Pasted image 20251106131327.png]]
+
+## advanced techniques for CNNs
+
+![[Pasted image 20251106131351.png]]
+
+
+![[Pasted image 20251106131407.png]]
+
+![[Pasted image 20251106131415.png]]
+![[Pasted image 20251106131447.png]]
+![[Pasted image 20251106131501.png]]
+
+
+
+![[Pasted image 20251106131514.png]]
+![[Pasted image 20251106131535.png]]
+![[Pasted image 20251106131549.png]]
+
+![[Pasted image 20251106131559.png]]
+
+Use a pre-trained model on a large dataset and 'adapt'iit to a task with a smaller dataset
+
+![[Pasted image 20251106131652.png]]
+3. freeze initial convolutional layers and fine-tune the last layers
+![[Pasted image 20251106131724.png]]
+
+## finetuning
+![[Pasted image 20251106131735.png]]
+
+conclusion:
+- images as input data
+- CNN architecture: ResNet
+- How to transfer learning to train the REsNet
+
+https://cs231n.github.io/transfer-learning/
+
+
+## things found during exercises:
+![[Pasted image 20251106132145.png]]
+
+
+
+![[Pasted image 20251106142417.png]]
+
+
+![[Pasted image 20251106142639.png]]
+
+
+
+Naive Bayes: missing feature values are: ignored!
+
+
+![[Pasted image 20251106145927.png]]
+
+![[Pasted image 20251106150042.png]]
+![[Pasted image 20251106150218.png]]
+
+![[Pasted image 20251106150412.png]]
+
+
+![[Pasted image 20251106150522.png]]
+(logistic regreession is linear, don't forget that - but you already kinda thought that right, after all the thing that goes in the sigmoid is linear)
+
+
+TODO: what does it mean for something to be separable
+![[Pasted image 20251106220110.png]]
+
+
+logistic regression simply has the linear thing in hte sigmoid function
+Model the class posterior probability $P(Y=1|X)$ using $\sigma(h(X))$
+![[Pasted image 20251106222245.png]]
+
+![[Pasted image 20251106222907.png]]
+
+
+![[Pasted image 20251106223635.png]]
+
+
+![[Pasted image 20251106224329.png]]
+parametric = you make an assumption on the distribution of the data
+
+![[Pasted image 20251106224411.png]]
+
+![[Pasted image 20251106224445.png]]
+![[Pasted image 20251106224648.png]]
+
+![[Pasted image 20251106225012.png]]
+
+
+![[Pasted image 20251105154154.png]]
+(don't forget kids - this is a logistic function, can be used as an activation function)
+
+![[Pasted image 20251106230015.png]]
+