Overview of application of deep learning in physics

Lecture 1

• Sylabus
• Some Remarks
• My contribution
• Machine learning: why and when?
• Machine learning: algorithms and tools

Lecture 2

• Scikit-learn package, its functionalities, in particular:
  • Perceptron
  • One, Two, Three, hidden neural networks for regression and classification problems

Lecture 3

• Mulilayer Perceptron
• Multilayer Perceptron in Scikit-Learn
  • Multilayer Perceptron in Scikit-Learn
    • Curve fitting
    • Classification
• Why is Machine Learning Difficult?

Lecture 4

• Tensor in Deep Learning
• PyTorch

Lecture 5

• AutoGrad
• Gradient Descent Algorithms
• Data analysis – simple example

Lecture 6

• optimization torch.optim
• torch.nn module
• linear and non-linear regression
• neural networks
• Mnist data
• DataSets, DataLoader
• Training with GPU

Lecture 7

• Convolution
• Pooling
• Convolutional Neural Networks
• MNIST-fashion data

Lecture 8

• MNIST-fashion by CNN
• Some examples of CNN’s
• LeNet and MNIST-fashion data
• Searching for the phase transition in the 2D Ising model

Lecture 9

• Overfitting - regularization
• DropOut
• Balanced Sampling

Lecture 10

• Bootstrapping
• BatchNormalization
• TensorFlow

Lecture 11

• Keras
• Eager Execution

Lecture 12

• TensorFlow: -tensors, variables -eager execution -automatic differentiation
• GANs

Lecture 13

• Practical implementations: calculating uncertianties.
  • DropOut
  • MeanVariance
  • Bootstraping