François Chung, Ph.D.

Tag: unsupervised learning

Deep learning and TensorFlow

Deep learning and TensorFlow

Cognitive Class training, MOOC (2020). This learning path presents the basic concepts of deep learning and TensorFlow with hands-on experience in solving problems. Throughout the training, TensorFlow is used in curve fitting, regression, classification and minimization of error functions. This concept is then explored in the deep learning world where TensorFlow is applied for backpropagation to tune the weights and biases while the neural networks are being trained.

Course 1: Deep learning fundamentals

Main topics:

  • Introduction to deep learning;
  • Deep learning models;
  • Additional deep learning models;
  • Deep learning platforms and libraries.

Course 2: Deep learning with TensorFlow

Main topics:

  • Introduction to TensorFlow;
  • CNN - Convolutional Neural Network;
  • RNN - Recurrent Neural Network;
  • Unsupervised learning.

References

Training

Deep learning fundamentals (course certificate)
Deep Learning Essentials (certification badge)
Deep learning with TensorFlow (course certificate)
Deep Learning using TensorFlow (certification badge)

Related articles

Learn more

Spark fundamentals

Spark fundamentals

Cognitive Class training, MOOC (2020). This learning path addresses the fundamentals of Apache Spark, an open source engine for large scale data processing that is revolutionizing the analytics and big data world. This training is an opportunity to learn from industry leaders about Spark, which is built around speed, ease of use and analytics, and provides hands-on opportunities and projects to build confidence with the Spark toolset.

Course 1: Spark fundamentals I

Main topics:

  • Introduction to Spark;
  • Resilient Distributed Dataset (RDD) and DataFrames;
  • Spark application programming;
  • Introduction to Spark libraries;
  • Spark configuration, monitoring and tuning.

Course 2: Spark fundamentals II

Main topics:

  • Introduction to notebooks;
  • RDD architecture;
  • Optimizing transformations and actions;
  • Caching and serialization;
  • Developing and testing.

Course 3: Spark MLlib

Main topics:

  • Spark MLlib data types;
  • Review of algorithms;
  • Decision trees and random forests;
  • Spark MLlib clustering.

Course 4: Exploring GraphX

Main topics:

  • Introduction to Graph-Parallel;
  • Exploring graph operators;
  • Visualizing and modifying GraphX;
  • Aggregation and caching.

Course 5: Big data in R using Spark

Main topics:

  • Introduction to SparkR;
  • Data manipulation in SparkR;
  • Machine learning in SparkR.

References

Training

Spark fundamentals I (course certificate)
Spark – Level 1 (certification badge)
Spark fundamentals II (course certificate)
Spark MLlib (course certificate)
Exploring GraphX (course certificate)
Big data in R using Spark (course certificate)
Spark - Level 2 (certification badge)

Related articles

Hadoop fundamentals (Cognitive Class training)
Data science specialization (Coursera training)

Learn more

Data science specialization

Data science specialization

Coursera training, MOOC (2020). This specialization covers the concepts and tools needed throughout the entire data science pipeline, from asking the right kinds of questions to making inferences and publishing results. Topics covered include using R to clean, analyze, and visualize data, navigating the entire data science pipeline from data acquisition to publication, using GitHub to manage data science projects, and performing regression analysis, least squares and inference using regression models.

Course 1: Data scientist’s toolbox

Main topics:

  • Data science fundamentals;
  • R and Rstudio;
  • Version control and GitHub;
  • R Markdown, scientific thinking and big data.

Course 2: R programming

Main topics:

  • Background and getting started;
  • Programming with R;
  • Loop functions and debugging;
  • Simulation and code profiling.

Course 3: Getting and cleaning data

Main topics:

  • Finding data and reading different file types;
  • Data storage systems;
  • Organizing, merging and managing data;
  • Text and data manipulation in R.

Course 4: Exploratory data analysis

Main topics:

  • Analytic graphics and base plotting in R;
  • Lattice and ggplot2;
  • Data dimension reduction;
  • Cluster analysis techniques.

Course 5: Reproducible research

Main topics:

  • Concepts, ideas and structure;
  • Markdown and knitr;
  • Reproducible research checklist;
  • Evidence-based data analysis.

Course 6: Statistical inference

Main topics:

  • Probability and expected values;
  • Variability, distribution and asymptote;
  • Intervals, testing and p-value;
  • Power, bootstrapping and permutation tests.

Course 7: Regression models

Main topics:

  • Least squares and linear regression;
  • Linear and multivariate regression;
  • Residuals and diagnostics;
  • Logistic and Poisson regression.

Course 8: Practical machine learning

Main topics:

  • Prediction, errors and cross validation;
  • Caret package;
  • Decision trees and random forests;
  • Regularized regression and combining predictors.

Course 9: Developing data products

Main topics:

  • Shiny, GoogleVis and Plotly;
  • R Markdown and Leaflet;
  • R Pakages and Swirl.

References

Related articles

Spark fundamentals (Cognitive Class training)
Hadoop fundamentals (Cognitive Class training)
AWS: foundations and machine learning (AWS training)

Learn more

CVIU 2013 - Academic journal

CVIU 2013 – Academic journal article

Publication

François Chung, Hervé Delingette; Regional appearance modeling based on the clustering of intensity profiles; In: Computer Vision and Image Understanding (CVIU), 117 (6), pp. 705-717, 2013.

Abstract

Model-based image segmentation is a popular approach for the segmentation of anatomical structures from medical images because it includes prior knowledge about the shape and appearance of structures of interest. This paper focuses on the formulation of a novel appearance prior that can cope with large variability between subjects, for instance due to the presence of pathologies. Instead of relying on Principal Component Analysis (PCA) such as in Statistical Appearance Models (SAMs), our approach relies on a multimodal intensity profile atlas from which a point may be assigned to several profile modes consisting of a mean profile and its covariance matrix. These profile modes are first estimated without any intra-subject registration through a boosted Expectation-Maximization (EM) classification based on spectral clustering. Then, they are projected on a reference mesh whose role is to store the appearance information in a common geometric representation. We show that this prior leads to better performance than the classical monomodal PCA approach while relying on fewer profile modes.

Keywords

  • appearance modeling
  • medical imaging
  • model-based image segmentation
  • unsupervised clustering

References

Publication

Related articles

3D Anatomical Human (INRIA project)
Ph.D. Thesis 2011 (doctoral thesis)

LAP 2011 - Book

LAP 2011 – Book

Publication

François Chung; Regional appearance modeling for model-based image segmentation: Methodological approaches to improve the accuracy of model-based image segmentation; Lambert Academic Publishing (LAP), Saarbrücken, 2011; ISBN: 978-3844322095.

Abstract

This thesis presents a novel appearance prior for model-based image segmentation. This appearance prior, denoted as Multimodal Prior Appearance Model (MPAM), is built upon an Expectation–Maximization (EM) clustering of intensity profiles with model order selection to automatically select the number of profile classes. Unlike classical approaches based on Principal Component Analysis (PCA), the clustering is considered as regional because intensity profiles are classified for each mesh and not for each vertex. Comparative results on liver profiles from Computed Tomography (CT) images show that MPAM outperforms PCA-based appearance models. Finally, methods for the analysis of lower limb structures from Magnetic Resonance (MR) images are presented. A first part deals with the creation of subject-specific models for kinematic simulations of the lower limbs. In a second part, the performance of statistical models is compared in the context of lower limb bone segmentation when only a small number of datasets is available for training.

References

Publication

Book (Amazon)
Book (MoreBooks)
Bibliographic reference (BibTeX)

Related articles

3D Anatomical Human (INRIA project)
Ph.D. Thesis 2011 (doctoral thesis)

Learn more

LAP – Lambert Academic Publishing