Advanced Machine Learning Masterclass
Advanced Machine Learning Masterclass
This course is for experienced machine-learning practitioners who are seeking to improve their skills and understanding of the field and to develop proficiency in building more accurate, efficient, and robust models. The aim of the course is to connect deeper theory to practice, so you can create faster, more accurate and appropriate models, and make more effective use of related techniques.
Ideal preparation for this course includes Presciient's course "Predictive Modelling, Data Science and Big Data," as well as work experience in the field, self-study, and participation in online predictive modelling tasks such as those offered by Coursera, Cloudera, and Kaggle. Participants should ideally be familiar with R, and have experience in using R for machine learning. Attendees should have some background in relevant statistics, as well as some practical experience with machine learning.
This course will cover advanced machine learning tools such as bagging, Lasso, elastic net, randomForest, gradient boosting, neural networks, and deep learning, giving students hands-on experience in using them.
The course will also cover key issues in modern machine learning: sparse data sets, and wide data sets with large numbers of categorical fields.
Matters vital to improved model accuracy, such as feature selection and feature generation, will be dealt with, along with methods to control overfitting in large data sets, including regularisation, dropout, and bagging.
The course will also provide experience with exploratory techniques for the investigation of large data sets, and discuss preprocessing techniques for atypical data sets such as network data.
Central to the course will be methods of error measurement and model selection, including k-fold cross-validation, out-of-time sampling, out-of-bag-based early stopping in boosting, regularisation, and advanced methods such as nested k-fold cross-validation. This section will also include a discussion of the theory of controlling overfitting, measuring model stability over time, and the benefits of robust, simple models. There will also be discussion of "the Curse of Dimensionality," which involves issues with high-dimensional spaces and time variation in the system being modelled.
As well as supervised machine learning, the course will present advanced unsupervised learning methods for data exploration and outlier detection. These will include randomForest-based metric-independent outlier detection and clustering, as well as neural autoencoding (the basic building block of deep learning and an automated method of feature selection) and methods such as principal components analysis and singular value decomposition.
Participants will be introduced to a range of tools in R for enabling advanced machine learning, including:
- Gbm for gradient boosting
- Glmnet for elastic net regularised generalised linear models
- Kernlab for support vector machines
In addition to tools in R, this course will time permitting, introduce students to Vowpal Wabbit, an extremely fast and scalable machine learning tool, and Theano, a GPU-enabled deep-learning library in Python, and the use of cloud-based tools to perform advanced machine learning.
Presciient training, coaching, mentoring, and capability development for analytics
Please ask about tailored, in-house training courses, coaching analytics teams, executive mentoring and strategic advice, and other services to build your organisation's strategic and operational analytics capability.
Our courses include:
- Introduction to R
- Predictive Modelling, Data Science and Big Data
- Forecasting and Trend Analysis
- Data Visualization
- Data Analytics for Fraud and Anomaly Detection in Forensics and Security
- Data Analytics for Campaign Marketing, Targeting and Insights
- Data Analytics for Insurance Claims analysis
- Data Analytics for Retail Marketing and Pricing
- Data Analytics for the Web
- Working with Data: Analysis and Report Writing for Everybody
I found the Introduction to R course extremely helpful. I have had very limited experience with R (and programming / statistical computing in general) and I now feel confident that I can use the language to do what I need with my data. The course was well designed and the notes are very helpful. I recommend this course to anyone who is new to R and wants to learn quickly.
- Helen McCormick, PhD student, Epigenetics Laboratory, Victor Chang Cardiac Research Institute
The Introduction to R course provided clear and logical assistance to getting up and running with R. More than that, the real value was in providing guidance on the myriad of online resources and introducing me to a network of passionate and helpful R users. Eugene is a knowledgeable and approachable teacher. I wouldn't hesitate in recommending the course. I feel that I am now fully on the road to applying R and using data to improve efficiency across my organisation.
- James Orton, Data and IT Manager, UNICEF Australia
I have been trying to convert my Stata programming skills to R, however, there have been many times where I just wanted to sit down with someone and have them explain the fundamentals of programming in R. Sure, a number of books and websites have helped me become familiar with R, however, I still didn't feel ready to translate all of my familiar Stata commands to R (e.g. I am comfortable plotting graphics using ggplot2, however, revert back to Stata for data manipulation). I knew that a more effective way to learn and feel confident would be to sit down with someone and have them explain how they use R, how they clean data, how they plot graphics, etc. I knew that once I felt comfortable with cleaning my data in R, analysis would be less of an issue - I'm happy to research the specifics on my own.
Thank you Eugene for advancing my R skills. I especially appreciate the time spent explaining the fundamentals of data manipulation - i.e. the code one needs to know before running any basic or sophisticated analysis. The pace of the workshop was perfect.
- Dr Chelsea Wise, Lecturer, Marketing, UTS Business School
Please ask about our discounts for group bookings.
Course material may vary from what is advertised due to the demands and learning pace of attendees. Additional material may be presented along with or in place of what is advertised.
The course may be cancelled by the organisers with full refund of fees up to a week in advance of the scheduled commencement date.