AutoMl vs ML Models

By Dr Mabrouka Abuhmida

AutoML (Automated Machine Learning) refers to the process of automating the entire machine learning process, including model selection, hyperparameter tuning, and feature engineering.

The goal of AutoML is to make it easier for non-expert users to build and deploy machine learning models without requiring knowledge of programming or advanced mathematical concepts. AutoML systems typically provide a user-friendly interface or API that allows users to specify the data they want to use to train a model, and the system takes care of the rest. Some popular AutoML systems include Google’s AutoML, H2O.ai’s AutoML, and DataRobot.

Definition:

“AutoML” refers to the process of automating the entire machine learning process, including model selection, hyperparameter tuning, and feature engineering. “ML models,” on the other hand, refer to the machine learning models that are built and used to make predictions or decisions based on data.

The concept of AutoML has been around since the early days of machine learning. Still, it has gained a lot of grip in recent years with of expansion of big data and the increasing demand for machine learning in various industries. Some of the earliest examples of AutoML systems include C4.5, an automatic decision tree learning algorithm developed by Ross Quinlan in the 1990s, and Auto-WEKA, a system for automatic model selection and hyperparameter optimization developed at the University of Waikato in New Zealand. However, the term “AutoML” itself only started to become widely used in the early 2010s, with the development of a number of commercial AutoML platforms, such as Google’s AutoML and H2O.ai’s AutoML.

AutoML is useful for a number of reasons. For one, it can help non-expert users build and deploy machine learning models without requiring knowledge of programming or advanced mathematical concepts. This can make it easier for organizations to adopt machine learning and start realizing the benefits of using it to make data-driven decisions.

AutoML is also useful because it can save time and effort for expert users. Building and tuning machine learning models can be a time-consuming process, especially when dealing with large and complex datasets. AutoML systems can automate many of the tasks that would otherwise be done manually, freeing up time for experts to focus on other aspects of the machine learning process, such as data preprocessing and feature engineering.

In contrast, ML models are the algorithms or mathematical models that are used to make predictions or decisions based on data. These models can be simple, such as linear regression or logistic regression, or more complex, such as support vector machines or neural networks. ML models are often trained on data and then used to make predictions or decisions on new data. Designing, Building, and Deploying ML models involve initialising and tuning hyperparameters, which in some cases could be in millions; for example, the YOLO model has over 14 million trainable hyperparameters, which do not include the input size and shape.

---

Here is an example of how AutoML could be used in a real-world scenario:

Scenario:

Suppose a company wants to use machine learning to predict customer churn (i.e., whether a customer is likely to leave the company). The company has a dataset containing information about its customers, such as their age, income, location, and whether they have churned in the past.

The company decides to use AutoML to build a machine learning model to predict customer churn. To do this, they use an AutoML platform to upload their dataset and specify that they want to build a model to predict churn. The AutoML platform then automatically selects the best model for the task, based on the characteristics of the data and the performance of various models on the task. It also automatically tunes the hyperparameters of the model to optimize its performance.

Once the model has been trained, the company can use it to make predictions on new data (i.e., data for customers that they have not seen before). For example, they might use the model to predict which of their current customers are most likely to churn in the future, and take steps to try to retain those customers.

This is just one example of how AutoML could be used in practice. There are many other potential use cases for AutoML, such as building models for fraud detection, predictive maintenance, or recommendation systems. The specifics of how AutoML is used will depend on the specific problem that the company is trying to solve and the characteristics of the available data.

---

It is difficult to predict exactly what the next generation of AutoML will look like, as it will depend on the advancements in machine learning and the specific needs of users. However, here are a few potential directions that AutoML could take in the future:

Improved automation: Future versions of AutoML may automate even more of the machine learning process, including tasks such as data preprocessing, feature engineering, and model deployment.

• Enhanced accuracy: AutoML systems may be able to produce more accurate models than humans by leveraging more advanced machine-learning techniques and larger amounts of data.

• Greater customization: AutoML systems may allow users to specify more granular preferences and constraints, such as the types of models they are interested in using or the trade-offs they are willing to make between accuracy and interpretability.

• Integration with other tools and platforms: AutoML systems may be more closely integrated with other tools and platforms, such as data visualisation tools or cloud computing platforms, to make it easier for users to work with machine learning in their workflow.

These are a few potential directions for the next generation of AutoML. AutoML will likely continue to evolve and improve over time in response to the needs and demands of users.

To summarize, AutoML is a process that helps automate the creation and deployment of ML models, while ML models are the algorithms or mathematical models used to make predictions or decisions based on data.