BIOF509 - Applied Machine Learning with Python
Machine learning is a computational field that consists of techniques allowing computers to learn from data and make data-driven predictions or decisions. The ability to implement machine learning approaches appropriately and intelligently is a crucial component of data analysis. BIOF 509 provides a broad practical introduction to machine learning concepts, analysis design, and implementation.
The course will give a broad and conceptual overview of the most popular machine learning algorithms, followed by examples of how and when to apply them to real data. Best practices in designing machine learning analyses will be emphasized and reviewed, along with how to avoid common pitfalls and how to interpret analysis results.
Through homework and in-class assignments, students will implement machine learning models in Python, utilizing state-of-the-art machine learning Python packages, such as scikit-learn and tensorflow. Algorithms that will be covered include but are not limited to linear and logistic regression, random forest, K-means clustering, and deep learning.
Note that the course emphasizes hands-on application of algorithms, and mathematical derivation will not be reviewed. Further, depending on the student’s familiarity with Python, completing the weekly homework assignments can take 1-4 hours.
The course will culminate in a short research project utilizing machine learning to analyze either the student’s own dataset or a public dataset that the student chooses.
- Ryan Dhindsa (email@example.com)
First class: 13st September 2018 at 5pm in building 10, room 2-3330.
Final classes: 13th and 20th December 2018
This document is subject to revision. Last revised 29th September 2018.
By the end of this course you should be able to:
- Create data analysis programs in Python language together with numpy, pandas, matplotlib in Jupyter environment
- Describe the common types of machine learning and deep learning tasks
- Implement simple machine learning algorithms, such as linear and logistic regression models and clustering utilizing numpy
- Convert a data set into a form suitable for use by machine learning algorithms
- Choose appropriate machine learning techniques for data analyses and interpret their results.
- Properly design machine learning analysis pipelines and avoid common pitfalls.
- Complete a short research project using machine learning.
This is a 14 week course starting on the 13th September 2018, and finishing on 13-14th December 2018. Classes will take place between 5:00pm and 7:00pm each Thursday in building 10, room 2-3330 within the FAES Academic Center.
Attendance in class is strongly recommended; however, we realize other commitments will occasionally prevent attendance. Class materials will generally be distributed over the course website.
Most classes will have hands-on tutorials and assignments. Both practice and graded assignments will generally be provided. Graded assignments should be submitted prior to the following class, please follow the deadlines specified on OKpy website. So that you can follow along during class bringing a laptop to each class is strongly encouraged.
- October 2018 - Last day to drop/withdraw
- November 2018 - Last day to change status (credit or audit)
Each student is encouraged to bring their own laptop to each class. For the course, we will use Python 3. Any up-to-date python installation should work, but you must be able to install packages. The Anaconda Scientific Python Distribution from Continuum Analytics will likely be the easiest approach to configuring python if you do not already have python installed. The Anaconda installer will automatically install many of the packages we will use during the course.
There is no required textbook for this course.
We will link to relevant online resources throughout the course.
If you would like a refresher on the basics, the following resources may be useful:
- Learn python the hard way (ebook freely available from the author) by Zed A. Shaw. A video course is also available.
- Think python (ebook freely available from the author) by Allen B. Downey.
Further material is included on the Week 1 page.
The following books cover some of the same material we will cover during the course. These books are not required, and presented solely as an alternative starting point covering the course objectives.
- The Elements of statistical learning (ebook freely available from the authors) by Trevor Hastie, Robert Tibshirani, and Jerome Firedman.
- Python Machine Learning by Sebastian Rashka. Release of a second edition is imminent, notebooks are already available
Assignments and Grading¶
The emphasis of the course is on learning and mastering the skills covered. It is our hope that everyone will be able to complete the assignments and project. If some of the material appears unclear please ask on the corresponding Slack channel for clarification.
The final project is 50% of the course, with the weekly assignments representing the remainder.
Weekly assignments will generally consist of multiple components. Unless otherwise specified, each component will be graded pass / fail. A component will be graded as “pass” if it runs and produces the expected results. The final grade will be equal to the percentage of components that are graded as “pass” out of all the assignment’s components.
The final project will consist of the following components:
1) Project documentation. Each project should have documentation clarifying its goal and functionality. The code itself should be well-documented, with comments spread out to aid understanding. Functions and classes should have docstrings describing their functionality, inputs and outputs.
2) Project code. The code should be well-organized and easy to read. It should also be written modularly, so that each part of code is reusable. The code should run and produce the correct output under different conditions. It should also have robust error checking.
3) Project presentation. Each student will present their project at the end of the semester. The idea here is to present the project’s goals, input, and output, preferably while showing snippets of code.
Project grades will be determined based on the components outlined above, with each component representing 33% of the project grade.
Course materials will be distributed on this website in the corresponding weekly sections.
Week 1 (13 September): Course overview. Introduction to machine learning topics
Week 2 (20 September): Data retrieval and visualization with numpy, pandas, scikit-learn
Week 3 (27 September): Data wrangling, preprocessing, and normalization
Week 4 (04 October): Supervised learning 1: Regression problems
Week 5 (11 October): Supervised learning 2: Overfitting, regularization, hyperparameter optimization, and cross-validation
Week 6 (18 October): Supervised learning 3: Classification problems
Week 7 (25 October): Unsupervised learning 1: Clustering
Week 8 (01 November): Unsupervised learning 2: Latent variable models
Week 9 (08 November): Unsupervised learning 3: Dimensionality reduction and feature selection
Week 10 (15 November): Introduction to deep learning methods
** No class on 22 November:** Thanksgiving
Week 11 (29 November): Practicum on deep learning
Week 12 (06 December): Implementing machine learning workflows. Common pitfalls and best practices.
Week 13 (13 December): Project presentations and feedback. Part I.
Week 14 (20 December): Project presentations and feedback. Part II.