As machine learning and deep learning are spreading into all aspects of our lives, a one-point lecture was held for beginner and entry-level developers who are not AI experts.

The first webinar, 'Introduction to Machine Learning & Deep Learning for Beginners', was recently released by e4ds EEWebinar. This webinar, hosted by e4ds Electronic Technology Research Institute, introduced AI terminology and basic concepts while conducting a contest on YOLO version 8 object detection at e4ds Make, and provided education for non-majors and beginners in AI.

The Introduction to Machine Learning & Deep Learning for Beginners webinar explores machine learning and deep learning, and explains machine learning and supervised learning, regression, learning, and loss functions, gradient descent, artificial neural networks, deep learning architecture, backpropagation, CNN, and NN structures.

In the first part of the webinar that day, the lecture focused on a general explanation of regression and classification models in supervised learning, loss functions, and gradient descent.

Artificial intelligence encompasses learning and reasoning, and machine learning is included within it. Machine learning uses data to learn data characteristics and patterns and predict future values based on them. Deep learning is a branch of machine learning and is a set of algorithms that learn through neural networks.

Machine learning includes supervised learning, unsupervised learning, and reinforcement learning, and Research Fellow Jeon Jong-cheon focused on supervised learning. Supervised learning is divided into regression models and classification models. A regression model predicts continuous output values for input values, and a classification model determines which category the input data belongs to.

The former head researcher added, “For example, regression models are suitable for developing models that predict test scores based on study time, and classification models are suitable for predicting whether a student will pass or fail a test based on study time.”

In regression models, finding the weights and biases that produce the smallest error is an important factor. Former Principal Researcher said, “The loss function is a function that calculates the error value as the difference between the correct answer and the predicted value. If possible, it is calculated as a square to make the error a larger value in order to maximize the learning effect in machine learning.”

This prevents the problem that the error has a larger value due to squaring, and that the sum of the errors becomes 0 when squaring is not performed, making effective learning possible in machine learning.

Next, the former head researcher explained the Logistic Regression algorithm used in binary classification models. For example, in finding the boundary point that determines whether or not a test is passed, the Logistic Regression algorithm finds the optimal straight line in the training data distribution and classifies it based on this.

This Logistic Regression algorithm is known to have high accuracy among classification algorithms and is used as a basic component in deep learning.

Meanwhile, the second part of the webinar, 'Introduction to Machine Learning & Deep Learning for Beginners', will be held live on the e4ds EEWebinar on the 6th.