User:Schober86/Distance metric learning

Distance metric learning is a current research topic in the field of machine learning. It's motivated by the search of a meaningful distance metric in a given input space. Distance metric learning tries to generate meaningful distance metrics automatically with machine learning algorithms. There have been several algorithms proposed so far.

Distance metric learning algorithms can be categorized into two major categories: supervised learning and unsupervised learning. In supervised distance metric learning the goal is to predict the distance between two input samples whereas in unsupervised distance metric learning an embedding in a lower dimensional spaced is seeked. Therefore, unsupervised distance metric learning is also called manifold learning. Applications of distance metric learning include the k-nearest neighbor algorithm, clustering and content-based image retrieval.

There is also a mathematical connection between kernel learning and distance metric learning.

Applications: Age estimation from face images

The estimation of human age from face images is an interesting problem in computer vision. As an important hint for human communication, facial images comprehend lots of useful information including gender, expression, age, pose, etc. Unfortunately, compared with other cognition problems, age estimation from face images is still very challenging. This is mainly due to the fact that, aging progress is influenced by not only personal gene but also many external factors. Physical condition, living style etc. may accelerate or slower aging process. Besides, since aging process is slow and with long duration, collecting sufficient data for training is a fairly strenuous work.

Human faces, as important visual cues, convey a significant amount of nonverbal information to facilitate the real-world human-to-human communication. As a result, the modern intelligent systems are expected to have the capability to accurately recognize and interpret human faces in real time. Facial attributes, such as identity, age, gender, expression, and ethnic origin, play a crucial role in real facial image analysis applications including multimedia communication, human computer interaction (HCI), and security. In such applications, various attributes can be estimated from a captured face image to infer the further system reactions. For example, if the user's age is estimated by a computer, an age specific human computer interaction (ASHCI) system may be developed for secure network/system access control. The ASHCI system ensures young kids have no access to internet pages with adult materials. A vending machine, secured by the ASHCI system, can refuse to sell alcohol or cigarettes to the underage people. In image and video retrieval, users could retrieve their photographs or videos by specifying a required age range. Ad-agency can find out what kind of scroll advertisements can attract the passengers (potential customers) in what age ranges using a latent computer vision system.