EVERY TUESDAY, 2:00PM-3:30PM PST STARTING SEPTEMBER 14TH THROUGH NOVEMBER 9TH (9 EPISODES)
Learn Live: Foundations of Data Science for Machine Learning
Starting Tuesday, September 14, 2021 (PST), join Jason DeBoever and Glenn Stephens live on Learn TV and explore this nine-part “Foundations of data science for machine learning” series. Each week, we will be walking through Learn modules and answering your questions live. From basic classical machine learning models to exploratory data analysis and customizing architectures, you’ll be guided by easy to digest conceptual content and interactive Jupyter notebooks and will learn about the underlying concepts as well as how to get into building models with the most common machine learning tools.
You will work through the following Microsoft Learn Modules with the expert presenters.
Learning Resources
-
Microsoft Learn Learning Path
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
-
Microsoft Learn Module
Our Presenters.
Jason DeBoever
Microsoft Senior Product Manager
TwitterGlenn Stephens
Microsoft Senior Content Developer
TwitterLearn Live Episodes
Introduction to machine learning: September 14 – Episode 01 (2:00pm PT – 3:30pm PT)
A high-level overview of machine learning for people with little or no knowledge of computer science and statistics. You’ll be introduced to some essential concepts, explore data, and interactively go through the machine learning life-cycle - using Python to train, save, and use a machine learning model like we would in the real world. In this episode, you will:
- Explore how machine learning differs from traditional software.
- Create and test a machine learning model.
- Load a model and use it with new datasets.
Build classical machine learning models with supervised learning: September 21 – Episode 02 (2:00pm PT – 3:30pm PT)
Supervised learning is a form of machine learning where an algorithm learns from examples of data. We progressively paint a picture of how supervised learning automatically generates a model that can make predictions about the real world. We also touch on how these models are tested, and difficulties that can arise when training them. In this episode, you will:
- Define supervised and unsupervised learning.
- Explore how cost functions affect the learning process.
- Discover how models are optimized by gradient descent.
- Experiment with learning rates, and see how they can affect training.
Introduction to data for machine learning: September 28 – Episode 03 (2:00pm PT – 3:30pm PT)
The power of machine learning models comes from the data that is used to train them. Through content and exercises, we explore how to understand your data, how to encode it so that the computer can interpret it properly, how to clean it of errors, and tips that will help you create models that perform well. In this episode, you will:
- Visualize large datasets with Exploratory Data Analysis (EDA).
- Clean a dataset of errors.
- Predict unknown values using numeric and categorical data.
Train and understand regression models in machine learning: October 5 – Episode 04 (2:00pm PT – 3:30pm PT)
Regression is arguably the most widely used machine learning technique, commonly underlying scientific discoveries, business planning, and stock market analytics. This learning material takes a dive into some common regression analyses, both simple and more complex, and provides some insight on how to assess model performance. In this episode, you will:
- Understand how regression works.
- Work with new algorithms: Linear regression, multiple linear regression, and polynomial regression.
- Understand the strengths and limitations of regression models.
- Visualize error and cost functions in linear regression.
- Understand basic evaluation metrics for regression.
Refine and test machine learning models: October 12 – Episode 05 (2:00pm PT – 3:30pm PT)
When we think of machine learning, we often focus on the training process. A small amount of preparation before this process can not only speed up and improve learning but also give us some confidence about how well our models will work when faced with data we have never seen before. In this episode, you will:
- Define feature normalization.
- Create and work with test datasets.
- Articulate how testing models can both improve and harm training.
Create and understand classification models in machine learning: October 19 – Episode 06 (2:00pm PT – 3:30pm PT)
Classification means assigning items into categories or can also be thought of automated decision making. Here we introduce classification models through logistic regression, providing you with a stepping-stone toward more complex and exciting classification methods. In this episode, you will:
- Discover how classification differs from classical regression.
- Build models that can perform classification tasks.
- Explore how to assess and improve classification models.
Select and customize architectures and hyperparameters using random forest: October 26 – Episode 07 (2:00pm PT – 3:30pm PT)
More complex models often can be manually customized to improve how effective they are. Through exercises and explanatory content, we explore how altering the architecture of more complex models can bring about more effective results. In this episode, you will:
- Discover new model types– decision trees and random forests.
- Learn how model architecture can affect performance.
- Practice working with hyperparameters to improve training effectiveness.
Confusion matrix and data imbalances: November 1 – Episode 08 (2:00pm PT – 3:30pm PT)
How do we know if a model is good or bad at classifying our data? The way that computers assess model performance sometimes can be difficult for us to comprehend or can over-simplify how the model will behave in the real world. To build models that work in a satisfactory way, we need to find intuitive ways to assess them, and understand how these metrics can bias our view. In this episode, you will:
- Assess performance of classification models.
- Review metrics to improve classification models.
- Mitigate performance issues from data imbalances.
Measure and optimize model performance with ROC and AUC: November 9 – Episode 09 (2:00pm PT – 3:30pm PT)
Receiver operator characteristic curves are a powerful way to assess and fine-tune trained classification models. We introduce and explain the utility of these curves through learning content and practical exercises. In this episode, you will:
- Understand how to create ROC curves.
- Explore how to assess and compare models using these curves.
- Practice fine-tuning a model using characteristics plotted on ROC curves.
