Machine Learning

​

Module 1: Machine Learning

 

• Definition of Machine Learning

• Machine Learning Methodology

  • Data Storage

  • Abstraction

  • Generalization

  • Assessment

• Supervised Learning

• Unsupervised Learning

• Reinforcement Learning

• deep learning

• Typology of Machine Learning algorithms

• Steps to Implement an Algorithm

  • information collection

  • Exploratory Analysis

  • Model Training

  • Model Evaluation

  • Model improvements

  • Machine Learning in consumer credit risk

• Machine Learning in credit scoring models

• Quantum Machine Learning

​

Module 2: EDA Exploratory Analysis

​

• Data typology

• transactional data

• Unstructured data embedded in text documents

• Social Media Data

• data sources

• Data review

• Target definition

• Time horizon of the target variable

• Sampling

  • Random Sampling

  • Stratified Sampling

  • Rebalanced Sampling

• Exploratory Analysis:

  • histograms

  • Q Q Plot

  • Moment analysis

  • boxplot

• Treatment of Missing values

  • Multivariate Imputation Model

• Advanced Outlier detection and treatment techniques

  • Univariate technique: winsorized and trimming

  • Multivariate Technique: Mahalanobis Distance

• ​Exercise 1: EDA Exploratory Analysis​

​

Module 3: Feature Engineering​

• Data Standardization

• Variable categorization

  • Equal Interval Binning

  • Equal Frequency Binning

  • Chi-Square Test

• binary coding

• Binning

  • Kind of transformation

  • Univariate Analysis with Target variable

  • Variable Selection

  • Treatment of Continuous Variables

  • Treatment of Categorical Variables

  • Gini

  • Information Value

  • Optimization of continuous variables

  • Optimization of categorical variables

• Exercise 2: Detection and treatment of Advanced Outliers

• Exercise 3: Stratified and Random Sampling in R

• Exercise 4: Multivariate imputation model

• Exercise 5: Univariate analysis in percentiles in R

• Exercise 6: Continuous variable optimal univariate analysis in Excel

• Exercise 7: Estimation of the KS, Gini and IV of each variable in Excel

• Exercise 8: Feature Engineering of variables

​

Unsupervised Learning

​

Module 4: Unsupervised models

​

• Hierarchical Clusters

• K Means

• standard algorithm

• Euclidean distance

• Principal Component Analysis (PCA)

• Advanced PCA Visualization

• Eigenvectors and Eigenvalues

• Exercise 9: Segmentation of the data with K-Means R

 

Supervised Learning

​

Module 5: Logistic Regression and LASSO Regression

 

• Econometric Models

  • Logit regression

  • probit regression

  • Piecewise Regression

  • survival models

• Machine Learning Models

  • Lasso Regression

  • Ridge Regression

• Model Risk in Logistic Regression

• Exercise 10: Lasso Logistic Regression in R

• Exercise 11: Ridge Regression in R

​

Module 6: Trees, KNN and Naive Bayes

 

• Decision Trees

  • modeling

  • Advantages and disadvantages

  • Recursion and Partitioning Processes

  • Recursive partitioning tree

  • Pruning Decision tree

  • Conditional inference tree

  • tree display

  • Measurement of decision tree prediction

  • CHAID model

  • Model C5.0

• K-Nearest Neighbors KNN

  • modeling

  • Advantages and disadvantages

  • Euclidean distance

  • Distance Manhattan

  • K value selection

• Probabilistic Model: Naive Bayes

  • Naive bayes

  • Bayes' theorem

  • Laplace estimator

  • Classification with Naive Bayes

  • Advantages and disadvantages

• Exercise 12: KNN and PCA

​

Module 7: Support Vector Machine SVM

​

• Support Vector Classification

• Support Vector Regression

• optimal hyperplane

• Support Vectors

• add costs

• Advantages and disadvantages

• SVM visualization

• Tuning SVM

• kernel trick

• Exercise 14: Support Vector Machine in R

​

Module 8: Ensemble Learning

​

• Classification and regression ensemble models

• bagging

• bagging trees

• Random Forest

• Boosting

• adaboost

• Gradient Boosting Trees

• xgboost

• Advantages and disadvantages

• Exercise 15:  Boosting in R

• Exercise 16: Bagging in R

• Exercise 17: Random Forest, R and Python

• Exercise 18:  Gradient Boosting Trees

​

​Deep Learning

​

Module 9: Introduction to Deep Learning

​

• Definition and concept of deep learning

• Why now the use of deep learning?

• Neural network architectures

• feedforward network

• R deep learning

• Python deep learning

• Convolutional Neural Networks

• Use of deep learning in image classification

• cost function

• Gradient descending optimization

• Use of deep learning

  • How many hidden layers?

  • How many neurons, 100, 1000?

  • How many times and size of the batch size?

  • What is the best activation function?

• Hardware, CPU, GPU and cloud environments

• Advantages and disadvantages of deep learning

 

Module 10: Deep Learning Feed Forward Neural Networks

​

• Single Layer Perceptron

• Multiple Layer Perceptron

• Neural network architectures

• activation function

  • sigmoidal

  • Rectified linear unit (Relu)

  • The U

  • Selu

  • hyperbolic hypertangent

  • Softmax

  • other

• Back propagation

  • Directional derivatives

  • gradients

  • Jacobians

  • Chain rule

  • Optimization and local and global minima

• Exercise 19: Deep Learning Feed Forward

​

Module 11: Deep Learning Convolutional Neural Networks CNN

​

• CNN for pictures

• Design and architectures

• convolution operation

• descending gradient

• filters

• strider

• padding

• Subsampling

• pooling

• fully connected

• Exercise 20: deep learning CNN

​

Module 12: Deep Learning Recurrent Neural Networks RNN

​

• Natural Language Processing

• Natural Language Processing (NLP) text classification

• Long Term Short Term Memory (LSTM)

• Hopfield

• Bidirectional associative memory

• descending gradient

• Global optimization methods

• One-way and two-way models

• Deep Bidirectional Transformers for Language Understanding ​

• Exercise 21: Deep Learning LSTM

​

Module 14: Generative Adversarial Networks (GANs)

 

• Generative Adversarial Networks (GANs)

• Fundamental components of the GANs

• GAN architectures

• Bidirectional GAN

• Training generative models

• Exercise 22: Deep Learning GANs

​

Module 15: Tuning Hyperparameters

​

• Hyperparameterization

• Grid search

• Random search

• Bayesian Optimization

• Train test split ratio

• Learning rate in optimization algorithms (e.g. gradient descent)

• Selection of optimization algorithm (e.g., gradient descent, stochastic gradient descent, or Adam optimizer)

• Activation function selection in a (nn) layer neural network (e.g. Sigmoid, ReLU, Tanh)

• Selection of loss, cost and custom function

• Number of hidden layers in an NN

• Number of activation units in each layer

• The drop-out rate in nn (dropout probability)

• Number of iterations (epochs) in training a nn

• Number of clusters in a clustering task

• Kernel or filter size in convolutional layers

• Pooling size

• Batch size

• Exercise 23: Optimization Xboosting, Random forest and SVM

• Exercise 24: Optimized Deep Learning

​

Probabilistic Machine Learning 

 

Module 16: Probabilistic Machine Learning

​​

• Introduction to probabilistic machine learning

• Gaussian models

• Bayesian Statistics

• Bayesian logistic regression

• Kernel family

• Gaussian processes

  • Gaussian processes for regression

• Hidden Markov Model

• Markov chain Monte Carlo (MCMC)

  • Metropolis Hastings algorithm

• Machine Learning Probabilistic Model

• Bayesian Boosting

• Bayesian Neural Networks

• Exercise 25: Gaussian process for regression

• Exercise 26: Bayesian Neural Networks

​

Model Validation

​

Module 17: Validation of traditional and Machine Learning models

​

• Model validation

• Validation of machine learning models

• Regulatory validation of machine learning models in Europe

• Out of Sample and Out of time validation

• Checking p-values in regressions

• R squared, MSE, MAD

• Waste diagnosis

• Goodness of Fit Test

• multicollinearity

• Binary case confusion matrix

• K-Fold Cross Validation

• Diagnostico del modelo

• Exercise 27: Validación avanzada de la regression

• Exercise 28: Diagnostico de la regresión

• Exercise 29: K-Fold Cross Validation in R

​

Module 18: Validación Avanzada de modelos de AI

 

• Integración de métodos de última generación en aprendizaje automático interpretable y diagnóstico de modelos.

• Data Pipeline

• Feature Selection

• Black-box Models

• Post-hoc Explainability

•  Global Explainability

• Local Explainability

• Interpretabilidad de Modelos

• Diagnóstico: Accuracy, WeakSpot, Overfit, Reliability, Robustness, Resilience, Fairness

• Comparativo de modelos

  • Comparativo para la Regresión y Clasificación

  • Fairness Comparison

• Ejercicio 30: Validación y diagnóstico de modelos avanzados de credit scoring

​​

 Auto Machine Learning and XAI 

​

Module 19: Automation of ML

 

• What is modeling automation?

• that is automated

• Automation of machine learning processes

• Optimizers and Evaluators

• Modeling Automation Workflow Components​

• Hyperparameter optimization

• Global evaluation of modeling automation

• Implementation of modeling automation in banking

• Technological requirements

• available tools

• Benefits and possible ROI estimation

• Main Issues

• Genetic algorithms

• Exercise 31: Automation of the modeling, optimization and validation of pricing models

​

Explainable Artificial Intelligence

​

Module 20: Explainable Artificial Intelligence XAI

 

• interpretability problem

• Machine learning models

  • 1. The challenge of interpreting the results,

  • 2. The challenge of ensuring that management functions adequately understand the models, and

  • 3. The challenge of justifying the results to supervisors

• ​Black Box Models vs. Transparent and Interpretable Algorithms

• interpretability tools

• Shap, Shapley Additive explanations

  • Global Explanations

  • Dependency Plot

  • Decision Plot

  • Local Explanations Waterfall Plot

• Lime, agnostic explanations of the local interpretable model

• Explainer Dashboard

• Other advanced tools

• Exercise 32: XAI interpretability of pricing

​

​

Quantum Computing

​

Module 21: Quantum computing and algorithms

  ​

Objective: Quantum computing applies quantum mechanical phenomena. On a small scale, physical matter exhibits properties of both particles and waves, and quantum computing takes advantage of this behavior using specialized hardware. The basic unit of information in quantum computing is the qubit, similar to the bit in traditional digital electronics. Unlike a classical bit, a qubit can exist in a superposition of its two "basic" states, meaning that it is in both states simultaneously.

​

• Future of quantum computing in insurance

• Is it necessary to know quantum mechanics?

• QIS Hardware and Apps

• quantum operations

• Qubit representation

• Measurement

• Overlap

• matrix multiplication

• Qubit operations

• Multiple Quantum Circuits

• Entanglement

• Deutsch Algorithm

• Quantum Fourier transform and search algorithms

• Hybrid quantum-classical algorithms

• Quantum annealing, simulation and optimization of algorithms

• Quantum machine learning algorithms

• Exercise 33: Quantum operations multi-exercises

​

Module 22: Introduction to quantum mechanics

​

• Quantum mechanical theory

• wave function

• Schrodinger's equation

• statistical interpretation

• Probability

• Standardization

• Impulse

• The uncertainty principle

• Mathematical Tools of Quantum Mechanics

• Hilbert space and wave functions

• The linear vector space

• Hilbert's space

• Dimension and bases of a Vector Space

• Integrable square functions: wave functions

• Dirac notation

• operators

• General definitions

• hermitian adjunct

• projection operators

• commutator algebra

• Uncertainty relationship between two operators

• Operator Functions

• Inverse and Unitary Operators

• Eigenvalues and Eigenvectors of an operator

• Infinitesimal and finite unit transformations

• Matrices and Wave Mechanics

• matrix mechanics

• Wave Mechanics

• Exercise 34: Quantum mechanics multi-exercises

​

Module 23: Introduction to quantum error correction

​

• Error correction

• From reversible classical error correction to simple quantum error correction

• The quantum error correction criterion

• The distance of a quantum error correction code

• Content of the quantum error correction criterion and the quantum Hamming bound criterion

• Digitization of quantum noise

• Classic linear codes

• Calderbank, Shor and Steane codes

• Stabilizer Quantum Error Correction Codes

• Exercise 35: Noise Model, Repetition Code and quantum circuit

​

​Module 24: Quantum Computing II

 

• Quantum programming

• Solution Providers

  • IBM Quantum Qiskit

  • Amazon Braket

  • PennyLane

  • cirq

  • Quantum Development Kit (QDK)

  • Quantum clouds

  • Microsoft Quantum

  • Qiskit

• Main Algorithms

  • Grover's algorithm

  • Deutsch–Jozsa algorithm

  • Fourier transform algorithm

  • Shor's algorithm

• Quantum annealers

• D-Wave implementation

• Qiskit Implementation

• Exercise 36: Quantum Circuits, Grover Algorithm Simulation, Fourier Transform and Shor​

​

Module 25: Quantum Machine Learning

​

• Quantum Machine Learning

• hybrid models

• Quantum Principal Component Analysis

• Q means vs. K means

• Variational Quantum Classifiers

• Variational quantum classifiers

• Quantum Neural Network

  • Quantum Convolutional Neural Network

  • Quantum Long Short Memory LSTM

• Quantum Support Vector Machine (QSVC)

• Exercise 37: Quantum Support Vector Machine​

 

​Module 26: Tensor Networks for Machine Learning

 

• What are tensor networks?

• Quantum Entanglement

• Tensor networks in machine learning

• Tensor networks in unsupervised models

• Tensor networks in SVM

• Tensor networks in NN

• NN tensioning

• Application of tensor networks in credit scoring models

• Exercise 38: Neural Network using tensor networks

 

Modelos de Healthcare

​

​Module 27: Detección temprana de la Obesidad usando AI y Quantum AI 

​​

La obesidad es una enfermedad epidémica, ya que el sobrepeso o la obesidad aumentan el riesgo de padecer enfermedades graves, como diabetes, cardiopatías, hipertensión y ciertos tipos de cáncer que conducen a la muerte prematura. Sin embargo, la identificación precoz de los factores causantes hace que la obesidad sea altamente prevenible. Debido al objetivo de la detección precoz y con el avance de los algoritmos de aprendizaje automático (ML), se evalúan algunos modelos empleados en la detección de la obesidad frente a algoritmos de ML cuántico. 

​

• Obesidad en en el mundo

• ¿Cuáles son las causas de la obesidad y el sobrepeso?

• ¿Cuáles son las consecuencias para la salud del sobrepeso y la obesidad?

• ¿Cómo reducir el sobrepeso y la obesidad?

• Variables relacionadas con los hábitos alimentarios

• Variables relacionadas con la condición física 

• Feature Engineering

• Importancia en la selección de variables

• Tratamiento de los outliers

• Modelos empleados en los últimos años para detectar la obesidad

• Modelos de Machine Learning

  • Support Vector Machine Regression

  • K-Nearest Neighborhood

  • Random Forests

  • Gradient Boosting

  • Extreme Gradient Boosting

• Quantum Machine Algorithms

  • Qubit and Quantum States

  • Quantum circuits

  • Support Vector Quantum Machine

  • Variational quantum classifier

  • Quantum Neural Networks

• Exercise 39: Random Forest, Gradient Boosting and Extreme Gradient Boosting to detect obesity

• Exercise 40: Quantum Support Vector Machine  and classical SVM to detect obesity

• Exercise 41: Quantum Neural Networks to detect obesity

• Exercise 42: Quantum Convoluted Neural Networks to detect obesity

​​

Module 28: Modelo de Retinopatía diabética con redes neuronales convolucionales

​

La retinopatía diabética es una de las principales causas de ceguera en personas diabéticas de entre 25 y 65 años. Las lesiones en la retina causadas por el debilitamiento de los vasos sanguíneos pueden provocar pérdida de visión e incluso ceguera total. Los métodos actuales de clasificación manual para detectar la retinopatía diabética requieren mucho tiempo y son propensos a errores. Las redes neuronales convolucionales han demostrado ser muy prometedoras para automatizar la identificación y categorización de la retinopatía diabética.​

​

• Los niveles excesivos de azúcar en sangre

• Estadísticos en el mundo sobre la diabetes 

• Diabetes-related retinopathy (DR)

• Cuatro fases

  • DR no proliferativa leve (RDNP)

  • DR no proliferativa moderada (RDNP)

  • Severe NPDR

  • Proliferative DR (PDR)

• El preprocesamiento de imágenes y proceso de diagnóstico de la retinopatía diabética mediante CNN

• Clasificación de imagenes en Real Estate

  • Problem Statement

  • Deep Learning Problem Formulation

  • Project and Data Source

  • Image Dataset

  • Evaluation Metric

  • Exploratory Data Analysis

  • Image Preprocessing

  • Model Training

  • Productionizing

• Image Classification: Data-driven Approach

• Convolutional Neural Networks

• Data and the Loss preprocessing,

• Hyperparameters

• Train/val/test splits

• L1/L2 distances

• Hyperparameter search

  • Optimization: Stochastic Gradient Descent

  • Optimization landscapes

  • Black Box Landscapes

  • Local search

  • Learning rate

• Weight initialization,

• Batch normalization

• Regularization (L2/dropout)

• Loss functions

• Gradient checks

• Sanity checks

• Hyperparameter optimization

•  Architectures

  • Convolution / Pooling Layers

  • Spatial arrangement,

  • Layer patterns,

  • Layer sizing patterns,

  • AlexNet/ZFNet/Densenet/VGGNet

• Convolutional Neural Networks tSNE embeddings

• Deconvnets

• Data gradients

• Fooling ConvNets  

• Human comparisons Transfer Learning and

• Fine-tuning Convolutional Neural Networks

• Performance Metrics

  • Accuracy

  • F1-Score

  • AUC-ROC

  • Cohen Kappa Coefficient

• Exercise 43: Convolutional Neural Networks para diagnóstico de la retinopatía diabética

​​

Module 29: Descubrimiento de fármacos usando

redes neuronales GAN

​

El descubrimiento de fármacos se refiere al proceso de identificación y desarrollo de nuevos compuestos químicos, para crear medicamentos que puedan tratar o curar enfermedades. En este proceso, uno de los mayores obstáculos es diseñar una molécula con las propiedades necesarias, ya que requiere numerosas optimizaciones químicas y estructurales. Las redes generativas adversariales pueden aprender la representación de las estructuras químicas y las propiedades de los fármacos del conjunto de datos para producir nuevas estructuras químicas que tengan propiedades similares al conjunto de entrenamiento. Las redes generativas adversariales tienen el potencial de acelerar el descubrimiento de fármacos mediante la generación de nuevos compuestos con propiedades deseables, reduciendo así el tiempo y el coste necesarios para los métodos tradicionales de descubrimiento de fármacos.

 

• Simplified Molecular Input Line Entry System (SMILES)

• Drug Discovery Cycle

• Generative Adversarial Networks (GANs)

• Backpropagation in Discriminator and Generator Training

• Encoding

• GAN Variants

  • CGAN

  • LAPGAN

  • DCGAN

  • AAE

  • INFOGAN

• Aplicación del GAN​

  • Descubrimiento de fármacos

  • Desarrollo de fármacos

  • Biomolecular

  • Targeting

• ​Modified and Applied CNN architectures used as the GAN generators and discriminators

• DCGAN generator and discriminator used for molecular compound fingerprint generation

• Exercise 44:  Modelos de descubrimiento de fármacos usando redes neuronales GAN

 

Quantum Computing for Healthcare

​

Module 30: Aplicaciones de computación cuántica en soluciones clinicas y médicas

​

La medicina, incluidas las ciencias de la salud y de la vida, ha sido testigo en los últimos años de un aluvión de actividades y experimentos relacionados con la computación cuántica. Al principio se centraron en problemas bioquímicos y de biología computacional, pero últimamente las soluciones cuánticas clínicas y médicas han despertado un interés creciente. La rápida aparición de la informática cuántica en los ámbitos de la salud y la medicina exige una adaptación del panorama.

​​

• Impacto de la computación cuántica en el healthcare

• Quantum support vector classifier (QSVC)

  • Virtual screening in drug discovery

• QNN, QSCV

  • Classification of ischemic heart disease

• Transfer learning-based QNN

  • Classification of breast cancer​​

• VQC

  • Classification of diabetes

• QSVC

  • Classification of medication persistence for individuals with rheumatoid arthritis

• Grover’s

  • ​DNA sequence alignment

• Medical quantum computing challenges

  • Data Security

  • Replicability

  • Skill devolpment

• Exercise 45: Hybrid Quantum Classical Neural Networks for Diabetic Retinopathy

• Exercise 46: Experimentos de diabetes usando Neural Network clásico y Quantum Neural Network 

• Exercise 47: Esophageal cancer tissue sorter using Neural Network clásico y Quantum Neural Network 

• Exercise 48: Quantum GAN para descubrimiento de fármacos 

​​

Probabilistic Machine Learning for Healthcare

​

Module 31: Bayesian Neural Network

​​​

 Los modelos de DL también se han utilizado intensamente en diferentes tareas de la atención sanitaria, como el diagnóstico y el tratamiento de enfermedades. Las técnicas de ML han superado a otros algoritmos de aprendizaje automático y han demostrado ser las herramientas definitivas para muchas aplicaciones de última generación. A pesar de todo ese éxito, el aprendizaje profundo clásico tiene limitaciones y sus modelos tienden a ser muy confiados en sus decisiones predichas porque no sabe cuándo se equivoca. Para el ámbito sanitario, esta limitación puede tener un impacto negativo en las predicciones de los modelos, ya que casi todas las decisiones relativas a pacientes y enfermedades son delicadas. Por ello, se ha desarrollado el aprendizaje profundo bayesiano (BDL) para superar estas limitaciones. A diferencia del DL clásico, el BDL utiliza distribuciones de probabilidad para los parámetros del modelo, lo que permite estimar todas las incertidumbres asociadas a los resultados predichos. En este sentido, el BDL ofrece un marco riguroso para cuantificar todas las fuentes de incertidumbre del modelo.

​

• Convolutional Neural Networks (CNN)

• Markov chain Monte Carlo (MCMC)

  • Metropolis Hastings algorithm

• Monte Carlo Dropout (MC–DROPOUT)

• Variational Inference (VI)

• Bayesian Neural Networks

• Exercise 49: Bayesian Neural Networks for Diabetic Retinopathy Detection 

​

​

​