You’ve probably wondered: can an algorithm be patented? If yes, then read on. You’ll discover the Enablement requirement for a patent application. Examples of allowable claims, and how algorithms can be patented. Then, you’ll learn about the different types of algorithms that are protected by patents. And at the end of the article, we’ll share some examples of algorithms that were patented.
Enablement requirement for a patent application
In a patent application for an algorithm, there is a requirement to meet what is called the “enablement requirement.” This term refers to the requirements under 35 U.S.C. 112(a) to define the claimed invention. A person of ordinary skill in the art is someone who can develop the invention without undue experimentation. Often a computer scientist or software engineer is the person to qualify as an ordinary skill in the art.
In order to meet the enablement requirement, an algorithm must be disclosed using working examples and/or prophetic examples. In general, a software “algorithm” can meet this requirement. These examples are required by 35 U.S.C. 112, and the patent application must include a working example of the algorithm. While the algorithm can be used to create a program, it must also be programmable.
When deciding whether an algorithm is sufficiently disclosed, a patent examiner may refer to the Section 112 Guidance to illustrate what constitutes “sufficient enablement.” For example, the patent in Sitrick v. Dreamworks, LLC involved a method for substituting the user’s audio signal and visual image into a video game. A patent applicant who fails to provide sufficient enablement can be found guilty of invalidation.
While citing excerpts from technical publications may satisfy the enablement requirement, a claim containing the same basic elements as the claimed device would lose patentability because it would not meet the enablement requirement. The claim must contain sufficient information to enable the skilled person to make the claimed invention. This burden, however, does not mean a patent applicant should abandon the invention. The applicant must prove that the prior art computer systems are sufficiently similar to the claimed device and that the circuits are interconnected to produce the required results.
In order to satisfy the enablement requirement, a patent examiner must identify the essential parts of the algorithm and the relationship between them. The enablement requirement should also bear a reasonable correlation to the claims. This means that an applicant must be able to demonstrate that there is at least one disclosed use of the algorithm. Once the claims are sufficiently clear, a patent examiner must examine the claims and explain why the enablement requirement is not satisfied.
The examiner must weigh all evidence presented in the patent application. This includes the specification, new evidence supplied by the applicant, and the examination of the previous rejection. Never base a decision on a personal opinion. Always consider the weight of all evidence in the record. So what is the Enablement Requirement for an Algorithm? Here’s What to Consider
Examples of allowable patent claims
Patent examiners typically don’t like two-line, short claims. In such cases, the patent examiner will likely reject the claim, even if the other independent claims are valid. However, this should be a relatively easy hurdle to overcome. During the patent application process, explain why the algorithm is necessary for the invention and provide examples of allowable patent claims for algorithms. A means-plus-function claim has the same scope as other claims, but is usually rejected.
The July Update provided an example of an algorithm that was patentable. The claim recited the steps necessary for relocating textual information in overlapping windows. While it was not specific to any technological environment, it was nonetheless a valid claim. Its novelty was based on a number of additional limitations that demonstrated how the algorithm improved the basic display function of the computer and the user interaction. The USPTO further explained how the additional limitations would transform the claim from a generic computer performing generic functions into an ordered combination.
Aside from these examples, you should be aware of the limitations of algorithm patent applications. In many cases, the algorithm may not be a novel invention, but the steps of the method should be outlined in the claim. The claim should focus on the most important steps of the algorithm and not on steps that have no real-world application. Any steps that are omitted should be put in dependent claims. Further examples of algorithm patents can be found on the USPTO’s website.
As with all abstract ideas, algorithms cannot be patented as natural phenomena or laws of nature. However, the legal system has identified several exceptions to this rule. In these cases, the algorithm must be tied to elements that amount to significantly more than the abstract idea. That way, the algorithm could be patentable subject matter. This rule isn’t absolute, but it is a good guideline to consider when writing an algorithm patent claim.
For an algorithm to be patentable, the claim must contain enough detail to allow the reviewer to evaluate the invention and determine whether it meets the necessary requirements. Moreover, the claim must not be so vague that the patent examiner won’t be able to distinguish it from existing solutions. Otherwise, the applicant’s claim will most likely be rejected. Even if the algorithm is not a new technology, it should be sufficiently covered by several existing algorithms, such as supervised, unsupervised, and reinforcement learning algorithms.
Another way to avoid a rejection is to make sure that the claimed algorithm contains all necessary steps to make it work. In most cases, patent lawyers will advise applicants to confine their algorithm claims to one step rather than trying to describe an entire process. Further, they will encourage applicants to tether their algorithms to modules so that the claim identifies a specific useful application. And it will be difficult to exclude such claims.
Examples of patents granted for algorithms
While algorithmic ideas are often considered abstract, courts have tended to confuse them with software and computer programs. This confusion can make it difficult to understand the issues surrounding the patenting of algorithms. Courts are required to make clear the conceptual relationship between algorithms and software. However, there are some notable exceptions to this rule. Examples of patents granted for algorithms include the following:
Algorithms that have a physical embodiment may be able to receive a patent. In such a case, the algorithm must be recognizable as a backbone. If an algorithm is applied to a mechanical device, it must be constructed by reference to each individual step. The steps must be carried out in the order specified by the algorithm. For this reason, algorithms can be patented if the steps are identifiable and observable.
The Lempel Ziv Welch algorithm, which is used in the GIF format, is an example of an algorithm. It was patented in 1985 by Unisys but the patent was not enforced and royalties were collected. The Luhn algorithm, which was first filed in 1954, validates a number. Today, it is used to verify credit card numbers, IMEI numbers, Canadian Social Insurance Numbers, and a variety of other number formats.
Another example of a patented algorithm is the Karmarkar algorithm, which was first patented in 1975. This algorithm was used to determine the frequency of aftershocks following a major earthquake. The algorithm has many uses, and the PredPol company has used it extensively. You can read more about the history of algorithms on OpenGenus’ website. So, how does a patent work? It’s important to understand the process before applying for a patent.
The USPTO issued two recent cases that illustrate the nuances of algorithms. The USPTO cites Example 39 in their October 2019 Patent Eligibility Guidance Update. Example 39 demonstrates a method for training a neural network to detect faces. The patent holder is able to address the issue of false positives by retraining the algorithm. The algorithm is then capable of recognizing faces. For example, the algorithm used by the AI company in a facial recognition system recognizes a person.
An algorithm may be essential to modern crystallographic research. If patented, it could halt advances in crystallography. However, it is important to understand that complex algorithms are fundamental theories to the crystallographic community. Without them, aircraft engineers cannot create new planes. So, strong patent enforcement is necessary to protect this field. However, if you’re a developer of such algorithms, it’s important to keep a few things in mind.
