The keys to QML patent success

On this co-published article, Haseltine Lake Kempner’s Laura Compton takes a sensible take a look at the right way to formulate claims and draft functions for quantum machine studying innovations in view of the EPO’s patent eligibility necessities

In quantum machine studying (QML), classical machine studying algorithms, or costly subroutines of them, are sometimes tailored to run on a quantum computing system. QML utilises quantum sources to enhance the execution time and/or the efficiency of classical machine studying algorithms.

Facets of QML that could be patentable embrace the utilisation of a quantum computing system to execute extra effectively all or a part of a classical machine studying algorithm (for instance, utilizing a quantum laptop to calculate classical distances extra effectively for nearest neighbour, kernel and clustering strategies), or to execute a mannequin itself (for instance, reformulating a stochastic mannequin as a quantum system). Different associated features embrace the reformulation of an optimisation drawback such that it might be solved utilizing a quantum computing system.

One other side of QML that could be patentable consists of enhancements to current QML algorithms or fashions (for instance, an enchancment that reduces the depth of the quantum circuit required to execute the algorithm or mannequin, and/or makes use of gates which might be much less advanced, and/or avoids repetition of sure subroutines of the algorithm). Some enhancements could also be particular to the issue being solved itself (for instance, modifying the operations utilized to a quantum computing system such {that a} extra restricted area of potential options to an optimisation drawback is then searched over by the system).

Innovations relating to those features will likely be thought of patentable subject material on the EPO when the quantum computing system is an integral a part of the invention.

For such innovations, the impartial claims are more likely to make some reference to the quantum computing system and the way during which the algorithm has been tailored to be carried out on it. The dependent claims, if not the impartial declare itself, ought to:

1. specify the preliminary state of the qubits of the quantum computing system;
2. the variables this preliminary state represents;
3. how the qubits are manipulated in accordance with the algorithm;
4. the output obtained by measurement; and
5. what the output represents.

In view of the EPO’s “technicality” necessities, having a dependent declare that specifies how the output of the quantum computing system and/or the output of the machine studying mannequin, is then utilized in some technical course of, is really helpful.

The place the invention pertains to extra normal QML strategies, or enhancements to such strategies (which might be utilized to a variety of issues throughout a variety of fields), additionally it is really helpful to supply a variety of totally different use instances that show how the invention could be utilized to totally different sensible issues within the dependent claims, or the outline,.

Quantum computing typically, in addition to QML, is a quickly evolving and complicated area. As such, drafting functions which meet the EPO’s sufficiency and readability necessities is usually a problem. Subsequently, when drafting patent specs, it’s best observe to incorporate a full mathematical description of the quantum implementation of the algorithm or mannequin, alongside how every operation being utilized to the qubits pertains to the algorithm or mannequin being carried out (for instance, describing how a collection of operations utilized to the qubits are consultant of an goal perform that’s to be minimised).

For innovations which relate to bettering current QML algorithms or fashions, detailed description on how the modifications to the quantum circuit allow the advance to be realised ought to be included. As with all quickly evolving area the place there’s a lack of universally accepted terminologies, for functions referring to quantum computing typically, the phrases used within the claims of the applying ought to be outlined within the description.

Lastly, experimental knowledge could be significantly helpful by way of demonstrating an enchancment in pace or accuracy over the prior artwork and could be helpful for supporting creative step arguments in later prosecution. It’s also price contemplating organising the technical drawback the invention solves by way of why classical processes endure from disadvantages that make them commercially or technically non-viable (for instance, too sluggish for actual time deployment).

To summarise, the factors above can be utilized to help in drafting QML innovations appropriate for submission to the EPO and can be utilized to supply the applicant with the absolute best probability of acquiring a commercially helpful patent.

Laura Compton is a patent legal professional within the Bristol workplaces of Haseltine Lake Kempner

Earlier articles by Haseltine Lake Kempner authors on this collection could be accessed right here:

The right way to safe AI patents in Europe

Drafting AI patent functions for fulfillment on the EPO – eligibility and declare formulation

Drafting AI patent functions for fulfillment on the EPO – drafting the total specification

Expertise traits – why patent your hidden AI?

Google and Samsung prime the checklist of candidates for AI-related patents on the EPO

The EPO and UKIPO approaches to AI and patentable subject material

How revised EPO pointers have an effect on remedy of AI innovations

Monetising knowledge, machine studying’s most dear asset