Daniel Haehn PRO
Hi, I am a biomedical imaging and visualization researcher who investigates how computational methods can accelerate biological and medical research.
Day 6
sudo apt-get update
sudo apt-get dist-upgrade -y
sudo do-release-upgrade
MNIST
Supervised Learning
LeNet-5 developed by Yann LeCun in 1998
Supervised Learning
9
Convolutional Neural Network
NUMBER_OF_CLASSES = 10
MNIST
Convolutional Neural Network
77°
0.856
1. Load Data
2. Setup Network
3. Train Network
4. Predict!
4 Steps
UNet by Ronneberger 2015
Regularization!
20%
Data Augmentation
Checkpointing
history = model.fit(X_train, y_train, validation_split=0.1, epochs=50,
callbacks=[callback_checkpoint])
saves the best model during training...
Early Stopping
Can also prevent overfitting!
1. Download the Oxford-IIIT Pet Dataset (choose 1 class only) or any other binary segmentation dataset
2. Setup the tiny UNet example for that data
3. Add Techniques from today (Dropout, Checkpointing, Early Stopping, Loss Plots)
JOURNAL
create a shared google doc to take notes!
IMPACT Day 6
By Daniel Haehn
IMPACT Day 6
Machine Psychology at UMass Boston. See https://mpsych.orghttps://mpsych.org
