In today’s world, Machine learning has become one of the popular and exciting fields of study. Machine learning models can now learn and more accurately predict the outcomes for even unseen data. The ideas in Machine learning overlap and receives from Artificial Intelligence and many other related technologies. Today, machine learning is evolved from Pattern Recognition and the concept that computers can learn without being explicitly programmed to perform specific tasks. We can use Machine Learning algorithms (e.g., Logistic Regression, Naive Bayes, etc.) to recognize spoken words, mine data, build applications that learn from data, and more. Moreover, the accuracy of these algorithms increases over time.
In this article, you will explore the differences between generative vs discriminative models, understand the nuances of generative model vs discriminative model, and discover examples of discriminative vs generative models in practical applications.
Learning Objectives
This article was published as a part of the Data Science Blogathon.
Machine learning models can be classified intotwo types: Discriminative and Generative. In simple words, a discriminative model makes predictions on unseen data based on conditional probability and can be used either for classification or regression problem statements. On the contrary, a generative model focuses on the distribution of a dataset to return a probability for a given example.
We, as humans, can adopt any of the two different approaches to machine learning models while learning an artificial language. These two models have not previously been explored in human learning. However, it is related to known effects of causal direction, classification vs. inference learning, and observational vs. feedback learning. So, In this article, our focus is on two types of machine learning models – Generative and Discriminative, and also see the importance, comparisons, and differences of these two models, considering aspects like generative and discriminative models.
Suppose we are working on a classification problem where our task is to decide if an email is spam or not spam based on the words present in a particular email. To solve this problem, we have a joint model over.
Therefore, the joint distribution of the model can be represented as
p(Y,X) = P(y,x1,x2…xn)
Now, our goal is to estimate the probability of spam email i.e., P(Y=1|X). Both generative and discriminative models can solve this problem but in different ways.
Let’s see why and how they are different!
In the case of generative models, to find the conditional probability P(Y|X), they estimate the priorprobability P(Y) and likelihood probability P(X|Y) with the help of the training data and use the Bayes Theorem to calculate the posterior probability P(Y |X):
In the case of discriminative models, to find the probability, they directly assume some functional form for P(Y|X) andthen estimate the parameters of P(Y|X) with the help of the training data.
The discriminative model refers to a class of models used in Statistical Classification, mainly used for supervised machine learning. These types of models are also known as conditional models since they learn the boundaries between classes or labels in a dataset.
Discriminative models focus on modeling the decision boundary between classes in a classification problem. The goal is to learn a function that maps inputs to binary outputs, indicating the class label of the input. Maximum likelihood estimation is often used to estimate the parameters of the discriminative model, such as the coefficients of a logistic regression model or the weights of a neural network.
Discriminative models (just as in the literal meaning) separate classes instead of modeling the conditional probability and don’t make any assumptions about the data points. But these models are not capable of generating new data points. Therefore, the ultimate objective of discriminative models is to separate one class from another.
If we have some outliers present in the dataset, discriminative models work better compared to generative models i.e., discriminative models are more robust to outliers. However, one major drawback of these models is the misclassification problem, i.e., wrongly classifying a data point.
Training discriminative classifiers or discriminant analysis involves estimating a function f: X -> Y, or probability P(Y|X)
Generative models are machine learning models that learn to generate new data samples similar to the training data they were trained on. They capture the underlying distribution of the data and can produce novel instances. Generative models find applications in image synthesis, data augmentation, and generating realistic content like images, music, and text.
Generative models are considered a class of statistical models that can generate new data instances. These models are used in unsupervised machine learning as a means to perform tasks such as:
Since these models often rely on the Bayes theorem to find the joint probability, generative models can tackle a more complex task than analogous discriminative models.
So, the Generative approach focuses on the distribution of individual classes in a dataset, and the learning algorithms tend to model the underlying patterns or distribution of the data points (e.g., gaussian). These models use the concept of joint probability and create instances where a given feature (x) or input and the desired output or label (y) exist simultaneously.
These models use probability estimates and likelihood to model data points and differentiate between different class labels present in a dataset. Unlike discriminative models, these models can also generate new data points.
However, they also have a major drawback – If there is a presence of outliers in the dataset, then it affects these types of models to a significant extent.
Image Source: medium.com
Training generative classifiers involve estimating a function f: X -> Y, or probability P(Y|X):
Let’s see some of the differences between the Generative vs Discriminative Models:
Aspect | Generative Models | Discriminative Models |
---|---|---|
Purpose | Model data distribution | Model conditional probability of labels given data |
Use Cases | Data generation, denoising, unsupervised learning | Classification, supervised learning tasks |
Common Examples | Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs) | Logistic Regression, Support Vector Machines, Deep Neural Networks |
Training Focus | Maximize likelihood of observed data, Capture data structure | Learn decision boundary, Differentiate between classes |
Example Task | Image generation, Inpainting (e.g., GANs, VAEs) | Text classification, Object detection (e.g., Deep Neural Networks) |
Now, lets look at the concrete differences between both the models:
Discriminative models draw boundaries in the data space, while generative models model how data distributes throughout the space. A generative model explains how the data generates, while a discriminative model focuses on predicting the data labels.
In mathematical terms, discriminative machine learning trains a model by learning parameters that maximize the conditional probability P(Y∣X). On the other hand, a generative model learns parameters by maximizing the joint probability of P(X, Y).
Discriminative models recognize existing data; they identify tags and sort data, allowing you to classify it, while generative models produce something new.
Since these models use different approaches to machine learning, both are suited for specific tasks i.e., Generative models are useful for unsupervised learning tasks. In contrast, discriminative models are useful for supervised learning tasks. GANs (Generative Adversarial Networks) as a competition between the generator, a component of the generative model, and the discriminator, highlighting the contrast between generative and discriminative models.
Generative models have more impact on outliers than discriminative models.
Discriminative models are computationally cheap as compared to generative models.
Let’s see some of the comparisons based on the following criteria between Generative vs Discriminative Models:
Generative models need fewer data to train compared with discriminative models since generative models are more biased as they make stronger assumptions, i.e., assumption of conditional independence.
In general, if we have missing data in our dataset, then Generative models can work with these missing data, while discriminative models can’t. This is because, in generative models, we can still estimate the posterior by marginalizing the unseen variables. However, discriminative models usually require all the features X to be observed.
If the assumption of conditional independence violates, then at that time, generative models are less accurate than discriminative models.
In conclusion, discriminative and generative models represent two basic approaches to machine learning that solve various tasks. The discriminative approach focuses on learning the decision boundary between classes, while generative models model the underlying data distribution. Understanding the difference between discriminative and generative models helps us to make better decisions about which approach to use for a particular task to build a more accurate machine-learning solution.
Hope you like the article! Generative vs discriminative models represent two fundamental approaches in machine learning. While generative models create new data by estimating the joint probability distribution, discriminative models focus on classifying data by estimating the conditional probability. Examples include Gaussian Mixture Models for generative and Support Vector Machines for discriminative models, highlighting their distinct applications in tasks like classification and data generation.
A. Discriminative models focus on modeling the decision boundary between classes, while probabilistic models focus on modeling the underlying probability distribution of the data.
A. Discriminative models support classification tasks, where the goal is to predict the class label of an input based on some features. They model the decision boundary between classes rather than modeling the distribution of the data.
A. Generative model examples include Variational Autoencoders (VAEs) for image generation and Generative Adversarial Networks (GANs) for creating realistic data like images and text.
A. No, CNN (Convolutional Neural Network) is not a generative model. It’s a type of neural network used mainly for tasks like image classification, not for generating data.
A. An example of a generative AI model is a language model like OpenAI’s GPT-3, which generates human-like text. A discriminative AI model example is logistic regression, which you use for binary classification tasks like spam detection.
The media shown in this article does not belong to Analytics Vidhya, and the author uses it at their discretion.