A Quick Recap of the MATLAB Machine Learning On ramp course

After spending few nights of revising the previous knowledge and learning new concepts, I am finally done with this Machine Learning course. This was truly a rewarding experience and looking forward to using the knowledge gained in practice.

In this article, I will give a quick recap of one might expect from the course as it might be possible that you are already familiar with some of the concepts.

There are four main modules covered in this short 2 hours self paced course.

• Classification Workflow
• Import and Preprocess Data
• Engineering Features
• Classification Models

Classification Workflow

In this module, we learnt how to build a simple model to perform a classification task. The small chunk size lessons also taught how to perform important operations in MATLAB like importing data from a file, preprocessing the data in the file such as prediction and calculating matching or mismatching rate, extract features and based on the information available, build a model. In the final section of the module, we learnt how to evaluate the performance of the model.

Import and Preprocess Data

The data from real life machine learning examples is huge and not just one file. For example, in the case of handwriting recognition to detect English alphabets, we must have a training dataset of 26 files with each consisting of a huge collection of images of handwritten alphabets. In this module, we learnt how to read multiple files and store them as a Datastore. Datastore is a repository for collections of data that are too large to fit in memory. A datastore allows you to read and process data stored in multiple files on a disk, a remote location, or a database as a single entity. It follows with the preprocessing of the data in the Datastore by performing the operations learnt in previous modules.

Engineering Features

In this module, we learnt how to calculate features from raw signals. This includes determining peaks, outliers and missing data in the dataset. Presence of such unwanted artifacts could result in incorrect statistical assessment of the training/testing dataset. The lessons in this module taught to calculate summary statistics, derivatives and correlations. The module concludes with the discussion and exercises on automating feature extraction task.

Classification Models

The module covered MATLAB’s Classification Learner application that enables users to explore, train, and evaluate different classification methods with minimal coding. The module lets a user explore different options, classification models, and other solutions offered by the classification learning applications. In addition to classification, we also learnt how to investigate misclassification, generate confusion matrices to identify false positives and false negatives and so on. Based on this information, we can improve our models. Such an approach is a common industry practice to make the AI/ML systems better and better.

Conclusion

This is a great investment of time both personally and professionally. The journey does not and should not end here, as one must be prepared to use the knowledge in many different fields. One can make use of open datasets freely available online. However, there are certain prerequisites and it is better if one already has some knowledge about mathematics and MATLAB.