Applied Machine Learning and AICornell

Certificate Program

Machine learning (ML) is the use and development of computer systems with the ability to learn and discover patterns in data. You even encounter some of these systems on a daily basis; for example, a computer program can determine whether an email is spam or not spam, and a computer program can find patterns among shoppers and recommend products tailored toward their needs and interests. Learning to analyze and visualize data in meaningful ways is a critical step in your study of ML.

In this course, you will start by exploring the role that machine learning plays in the industry for decision making and its impact on your role. The characteristics of a particular problem, the data you have to work with, and the questions you want to answer will dictate what type of ML approach, method, and algorithm needs to be used. Once you cover the basic role of machine learning and the process from start to finish, you will gain experience in industry-relevant tools such as Jupyter Notebooks, NumPy, and Pandas.

One of the most important steps in the machine learning process is understanding and preparing data. Before you can learn to train models, you need to ensure the data selected for your model is appropriate to solve the problem.

In this course, you will focus on taking raw data, analyzing and organizing it, and preparing it for the next stage of the machine learning process: modeling. You will practice identifying examples, along with their features and labels, to prepare for supervised learning. You will also practice organizing your data into a data matrix. You will learn about feature engineering, which will allow you to transform your data into a format that is most appropriate for your specific model. By the end of the course, you will be set up with the necessary foundations for managing data in ML.

You are required to have completed the following courses or have equivalent experience before taking this course:

• Machine Learning Foundations

After data has been prepared, the next step in the machine learning lifecycle is model training and evaluation. In this course, you will focus on the model training and evaluation process for supervised learning models and explore a few supervised learning algorithms that are commonly used. You will be introduced to the model training for two popular supervised learning algorithms: k-nearest neighbors (KNN) and decision trees (DT), exploring their applicability to classification problems. You will practice creating your own machine learning models using a popular Python package for machine learning called scikit-learn. By the end of this course, you will have new, applicable skills in training common ML models.

You are required to have completed the following courses or have equivalent experience before taking this course:

• Machine Learning Foundations
• Managing Data in Machine Learning

Linear models are a class of supervised learning models that are represented by an equation and use a linear combination of features and weights to compute the label of an unlabeled example. Linear models are simple to implement, fast to train, and relatively low in complexity.

In this course, you will explore several linear models, including logistic regression, one of the most powerful linear models used in classification. Logistic regression is used to predict the probability of an outcome. While the focus of the unit will be on logistic regression, you will also be introduced to a common linear model used to solve regression problems: linear regression. You will delve into important concepts specific to the training of linear models, including the optimization algorithm, gradient descent, and the loss function evaluation tool. You will be given the opportunity to implement a logistic regression model from scratch using NumPy, and you will see a demonstration of how a linear regression model can be used to solve real-world regression problems, applying your experience to relevant scenarios.

You are required to have completed the following courses or have equivalent experience before taking this course:

• Machine Learning Foundations
• Managing Data in Machine Learning
• Training Common Machine Learning Models

Once you have trained your model, how do you know whether it will generalize well to new data? In this course, you will focus on techniques that can be used to properly evaluate and improve a model's performance with the view toward producing the best model for your data and machine learning problem. You will explore different model selection methods that are used to find the best-performing model, and you will apply common out-of-sample validation methods that are used to test your model on unseen data in support of model selection.

You will also discover how both hyperparameter configurations as well as feature combinations play roles in model performance. Using your own implementation along with built-in scikit-learn libraries, you will determine the optimal hyperparameter configuration for your model and perform feature selection techniques to find the combination of features that results in the best model performance.

You are required to have completed the following courses or have equivalent experience before taking this course:

• Machine Learning Foundations
• Managing Data in Machine Learning
• Training Common Machine Learning Models
• Training Linear Models

Ensemble modeling is a helpful and important technique used in machine learning. It's a powerful approach to train multiple models and quantify them into a single prediction. There are three commonly used ensemble techniques: stacking, bagging, and boosting. So how do you know which ensemble method to use and when to use it?

In this course, you will explore stacking, bagging, and boosting techniques, including the motivation behind using each and understanding their optimal scenarios as well as their tradeoffs. By the end of this course, you will have observed a number of robust algorithm case studies, such as random forests and gradient boosted decision trees, that employ these methods. You will also have the opportunity to put this new knowledge into action by practicing building and optimizing various ensemble models.

You are required to have completed the following courses or have equivalent experience before taking this course:

• Machine Learning Foundations
• Managing Data in Machine Learning
• Training Common Machine Learning Models
• Training Linear Models
• Evaluating and Improving Your Model

Natural language processing (NLP) is a branch of artificial intelligence that helps machines process and understand human language in speech and text form. In order for machine learning models to process words and blocks of text, the text must first be transformed into numerical features. There are various NLP preprocessing techniques that accomplish this.

In this course, you will explore these techniques and the typical workflow for converting text data for NLP. You will also use a special scikit-learn utility that allows you to automate the workflow as a pipeline. At the end of the course, you will have the opportunity to explore neural networks, powerful ML models that are heavily used in the field of NLP. You will also discover different Python packages used to construct neural networks and see how to implement a feedforward neural network using Keras. You will then delve into deep neural networks, which are used to solve large-scale complex problems, and you will implement a deep neural network for sentiment analysis. By the end of this course, you will have a foundation in using ML for text analysis relevant to limitless real-life applications.

You are required to have completed the following courses or have equivalent experience before taking this course:

• Machine Learning Foundations
• Managing Data in Machine Learning
• Training Common Machine Learning Models
• Training Linear Models
• Evaluating and Improving Your Model
• Improving Performance With Ensemble Methods