Hands-on supervised machine learning

5 dezembro 2018, 09:00 • André David

• Supervised machine learning
• 1-D vs Nx1-D vs N-D
• Decision trees
• Ensembling: bagging and boosting
• Hyperparameters
• Over-fitting and under-fitting
• Training and cross-validation
• Neural networks
• Hinge activation and the DNN revolution
• Hands-on RAMP with iris dataset
• Sandbox code
• Public leaderboard (training dataset)
• Private leaderboard (test dataset)
• Over-fitting in real life

Data visualisation and machine learning

30 novembro 2018, 15:30 • André David

• Data visualisation
• Cognition vs perception
• Average and standard deviation meet the Anscombe quartet
• Attributes: quantitative, ordinal, categorical
• The pop-out effect
• The use of color
• The probability integral transform
• Pair plots
• Correlations vs mutual information
• Data pre-processing
• Range normalisation
• Gaussianisation
• PCA and ICA
• Machine learning
• Approximate median significance as figure-of-merit in HEP
• ROC curve: AUC and working point
• Classification, regression, clustering, and dimensionality reduction

Intro and ML in the Higgs to diphoton discovery

28 novembro 2018, 09:00 • André David

• Fundamentals
• Theory and experiment in particle physics
• Wason selection task
• Uncertainty
• Error and uncertainty - accuracy and precision
• Trials factor and look-elsewhere effect
• The discovery of the Higgs boson
• Open questions in particle physics
• MVAs in the Higgs diphoton discovery
• Energy regression and energy resolution
• Photon identification MVA
• Vertex choice MVA and probability of correct choice MVA
• Diphoton MVA - bringing together multiple black boxes

Métodos de Monte-Carlo

23 novembro 2018, 15:30 • José Luís Martins

Método de Monte-Carlo,  Algoritmo de Metropolis. Métodos para acelerar a simulação da têmpera em Monte-Carlo.

Modelos numéricos em Estrutura Electrónica III.  Dinâmica molecular.

21 novembro 2018, 09:00 • José Luís Martins