What the unsupervised learningcould deliver us (or, what not)

초록

The unsupervised learning, as a mode of machine learning, shall make a machine so clever that it recognizes, for instance, (the pictures of) cats among other sort of things. It is remarkable that there is no need to feed the machine explicit information about the category-specific properties of the cat in order to promote the machine to such level of cat-discerning competence. Though it is not known yet, how far this kind of competence would reach, it gives us certain hope for an objective classification, i.e. for one that is free from prejudices or cultural biases that infects us, human observers. Utilizing its power we might also be able to overcome the problem of theoryladenness of observation, whereby the Baconian ideal of observation shall be realized. In this talk I will present an evaluation of this prospect, investigating the process and structure of unsupervised learning.