Latent variables in the context of "Novum Organum"

In research design, especially in psychology, social sciences, life sciences and physics, operationalization (or operationalisation) is a process of defining the measurement of a phenomenon which is not directly measurable, though its existence is inferred from other phenomena. Operationalization thus defines a fuzzy concept so as to make it clearly distinguishable, measurable, and understandable by empirical observation. In a broader sense, it defines the extension of a concept—describing what is and is not an instance of that concept. For example, in medicine, the phenomenon of health might be operationalized by one or more indicators like body mass index or tobacco smoking. As another example, in visual processing the presence of a certain object in the environment could be inferred by measuring specific features of the light it reflects. In these examples, the phenomena are difficult to directly observe and measure because they are general/abstract (as in the example of health) or they are latent (as in the example of the object). Operationalization helps infer the existence, and some elements of the extension, of the phenomena of interest by means of some observable and measurable effects they have.

Sometimes multiple or competing alternative operationalizations for the same phenomenon are available. Repeating the analysis with one operationalization after the other can determine whether the results are affected by different operationalizations. This is called checking robustness. If the results are (substantially) unchanged, the results are said to be robust against certain alternative operationalizations of the checked variables.

In machine learning, a deep belief network (DBN) is a generative graphical model, or alternatively a class of deep neural network, composed of multiple layers of latent variables ("hidden units"), with connections between the layers but not between units within each layer.

When trained without supervision on a set of examples, a DBN can learn to probabilistically reconstruct its inputs. The layers then act as feature detectors. After this learning step, a DBN can be further trained with supervision to perform classification.