3.3.1 Artificial intelligence and machine learning 

Artificial intelligence and machine learning are based on computational models and algorithms for classification, clustering, regression and dimensionality reduction, such as artificial neural networks, genetic algorithms, support vector machines, k-means, kernel regression and discriminant analysis. Such computational models and algorithms are per se of an abstract mathematical nature, irrespective of whether they can be "trained" using training data. However, their use does not by itself render inventions related to artificial intelligence or machine learning non-patentable, and. Hence, the guidance provided in G‑II, 3.3 generally applies to these computational models and algorithms too. This means that, if a claim of an invention related to artificial intelligence or machine learning is directed either to a method involving the use of technical means (e.g. a computer) or to a device, its subject-matter has technical character as a whole and is thus not excluded from patentability under Art. 52(2) or Art. 52(3). In such cases, the computational models and algorithms themselves contribute to the technical character of the invention if they contribute to a technical solution to a technical problem, for example by being applied in a field of technology and/or by being adapted to a specific technical implementation.

Terms such as "support vector machine", "reasoning engine" or "neural network" may, depending on the context, merely refer to abstract models or algorithms and so do not, on their own, necessarily imply the use of a technical means. This has to be taken into account when examining whether the claimed subject-matter has technical character as a whole (Art. 52(1), Art. 52(2) and Art. 52(3)).

Artificial intelligence and machine learning can be applied in various fields of technology. For example, using a neural network in a heart monitoring apparatus to identify irregular heartbeats makes a technical contribution. The classification of digital images, videos, audio or speech signals based on low-level features (e.g. edges or pixel attributes for images) is another typical technical application of classification algorithms. More examples of technical purposes for which artificial intelligence and machine learning could be used are listed in G‑II, 3.3.

However, classifying text documents solely according to their textual content is not considered to be per se a technical purpose but a linguistic one (T 1358/09). Similarly, classifying abstract data records or even "telecommunication network data records" without any indication that a technical use is made of the resulting classification is per se not a technical purpose, even if the classification algorithm can be considered to have valuable mathematical properties such as robustness (T 1784/06).

Where a classification method serves a technical purpose, the steps of generating the training set and training the classifier may also contribute to the invention's technical character if they help to achieve that technical purpose.

The technical effect that a machine learning algorithm achieves may be readily apparent or established by explanations, mathematical proof, experimental data or the like. Mere assertions are not enough, but comprehensive proof is not required either. If the technical effect depends on particular characteristics of the training dataset used, the characteristics required to reproduce the technical effect must be disclosed unless the skilled person can determine them without undue burden using common general knowledge. However, in general, there is no need to disclose the specific training dataset itself (see also F‑III, 3 and G‑VII, 5.2).