Last week, we explored the realm of fake image generation using Generative Adversarial Networks (GANs), diving into the world of statistical distributions, and understanding how GANs employ competing neural networks to create realistic images. In this week’s lesson, we will shift our focus to different machine learning models and their applications across a variety of tasks and industries.

Classes of Models – Recap

To recap, machine learning models can be broadly classified into supervised and unsupervised learning. In supervised learning, the most common tasks are classification and regression. On the other hand, unsupervised learning focuses on tasks such as feature extraction and clustering.

Machine Learning Pipeline Revisited

It is important to remember that the machine learning pipeline remains consistent, regardless of the task at hand. You choose the model based on the task, but the flow remains the same: data acquisition, data preparation, model training, and testing.

Supervised Learning – Classification

In classification, the goal is to predict the class a new piece of data belongs to, such as identifying whether an image contains a cat or a dog. The target label is categorical, and the model needs to classify data accordingly. Neural Networks and Convolutional Neural Networks (CNNs) are commonly used for image classification, while Naïve Bayes is popular for spam detection. Logistic regression is used for binary outcomes, with probabilities rounded to 0 or 1 for classification purposes.

Other common models

Key Takeaways:

  • Classification predicts data classes
  • Common models: NNs, CNNs, Naïve Bayes, Logistic Regression

Supervised Learning – Regression

In regression, the goal is to predict a numerical value based on input data, such as estimating an individual’s weight based on their height. The target label is numerical, and the model needs to perform regression. Neural networks can handle various tasks, while linear regression and polynomial regression models are also frequently used. Logistic regression can be applied to binary probability prediction without rounding the probability.

Key Takeaways:

  • Regression predicts numerical values
  • Common models: Neural Networks, Linear Regression, Polynomial Regression, Logistic Regression

Unsupervised Learning – Feature Extraction

Feature extraction in unsupervised learning helps models piece information together for tasks like classification or clustering. Data can be clustered based on similarities between learned features. This task is unsupervised as the model learns the distribution of input data, whereas supervised models learn the probability of the output given the input.

Key Takeaways:

  • Feature extraction learns data features
  • Useful for classification and clustering

Unsupervised Learning – Clustering

In clustering, data is grouped without labels to understand underlying behavior. One major application is data augmentation or semi-supervised training, where data points are grouped into clusters based on similarities between features, and labels are assigned for supervised training later. This process can also help with dimensionality reduction.

Key Takeaways:

  • Clustering groups data based on similarities
  • Applications: data augmentation, dimensionality reduction


This week’s lesson provided a comprehensive overview of various machine learning models and their applications across different tasks and industries. We revisited the machine learning pipeline, delved into supervised learning for classification and regression tasks, and explored unsupervised learning for feature extraction and clustering. Understanding these different models and their applications is crucial for selecting the most suitable approach when tackling real-world problems.

As we move forward to next week’s lesson, we will focus on data processing practices, which play a critical role in the success of any machine learning project. We will explore techniques and best practices for data acquisition, cleaning, transformation, and feature engineering, providing you with valuable insights into the practical aspects of data processing. Stay tuned for an in-depth look at data processing and its impact on machine learning projects and outcomes!