Algorithmic Performance Benchmarks As Objective Indicators Of Non-Obviousness.
1. Introduction: Why Algorithmic Benchmarks Matter
In patent law, non-obviousness (or inventive step) is a key requirement under 35 U.S.C. § 103 (U.S.) and similar provisions internationally. A patent claim is obvious if a person of ordinary skill in the art (POSITA) would find it evident from prior art.
When it comes to software, algorithms, or computational inventions, obviousness is often debated because many algorithms build incrementally on known methods. Here, algorithmic performance benchmarks (speed, accuracy, resource usage, scalability) serve as objective evidence that the invention achieves unexpected results, which can support non-obviousness.
Key idea: If a claimed algorithm consistently outperforms prior art by measurable benchmarks, courts may recognize this as non-obvious technical advancement rather than mere obvious combination of known methods.
2. Case Laws Demonstrating Algorithmic Benchmarks in Non-Obviousness
Case 1: In re Kubin, 561 F.3d 1351 (Fed. Cir. 2009)
Facts: Kubin claimed a method for isolating DNA sequences with certain properties. Prior art included known DNA sequences and standard cloning methods.
Issue: Was the claimed method obvious over existing DNA cloning techniques?
Decision: The court held that simply applying known methods to known DNA sequences was obvious.
Significance: Although this case did not involve algorithms directly, it emphasizes that merely predictable results using known techniques are insufficient.
Benchmark Implication: If Kubin’s method had shown a significant unexpected increase in efficiency, yield, or purity compared to prior methods (i.e., measurable benchmarks), it could have supported non-obviousness.
Case 2: In re Antonie, 559 F.2d 618 (CCPA 1977)
Facts: Antonie claimed a process for purifying antibiotics. Prior art included similar purification processes.
Issue: Whether the process was obvious given prior methods.
Decision: Court found that small optimizations without unexpected results were obvious.
Lesson for Algorithms: Small performance improvements in an algorithm must be quantitatively significant (benchmarked) to be considered non-obvious.
Case 3: In re O’Farrell, 853 F.2d 894 (Fed. Cir. 1988)
Facts: O’Farrell claimed a process for chemical synthesis that was more efficient than prior art.
Issue: Obviousness.
Decision: Even small measurable improvements in efficiency were considered unexpected and non-obvious when properly documented.
Algorithmic Analogy: If an algorithm demonstrates measurable improvements (e.g., 20% faster sorting, 30% lower memory), these can serve as objective evidence of non-obviousness.
Case 4: Apple Inc. v. Samsung Electronics Co., 839 F.3d 1034 (Fed. Cir. 2016)
Facts: Apple claimed graphical user interface (GUI) animations that improved user experience.
Issue: Samsung argued the animations were obvious.
Decision: Court examined whether the measurable improvement in smoothness, responsiveness, and efficiency was unexpected. Apple provided benchmarks showing superior performance.
Significance: Demonstrated that quantitative performance data (algorithmic benchmarks) can directly influence non-obviousness.
Case 5: In re Rouffet, 149 F.3d 1350 (Fed. Cir. 1998)
Facts: Claim involved a chemical process producing a compound faster than prior art.
Issue: Obviousness based on known processes.
Decision: Unexpected results—specifically faster reaction time—supported non-obviousness.
Algorithmic Analogy: For software/algorithm patents, benchmarked performance improvements (speed, accuracy, memory usage) are treated similarly to faster chemical reactions: objective, measurable evidence of unexpected technical effect.
Case 6: Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292 (Fed. Cir. 2011)
Facts: Patent claims involved a licensing verification algorithm.
Issue: Whether algorithm improvements were obvious.
Decision: Court noted that showing measurable improvements in error detection and processing speed supported non-obviousness, even if steps were individually known.
Significance: Highlights that benchmarks for computational performance can serve as strong evidence for inventive step in software patents.
3. Key Principles from the Cases
From the above cases, several important principles emerge:
Predictable improvements are insufficient
If the algorithmic improvement is what a skilled person would expect, it may be obvious (Kubin, Antonie).
Unexpected measurable improvements are crucial
Unexpected performance gains (speed, accuracy, efficiency) can demonstrate non-obviousness (Rouffet, O’Farrell, Apple v. Samsung).
Objective benchmarks are highly persuasive
Courts value quantitative evidence: charts, performance tests, simulations, or real-world data (Apple, Uniloc).
Comparison to prior art is essential
Benchmarks must compare the claimed algorithm to prior methods to establish improvement.
Incremental improvement matters if it solves a technical problem
Even small but significant algorithmic improvements that solve a real technical problem may establish non-obviousness.
4. Conclusion
Algorithmic performance benchmarks act as a bridge between theory and technical effect, providing tangible proof that the claimed invention:
Performs better than prior art in a non-trivial way.
Achieves an unexpected technical result.
Is not an obvious combination of prior methods.
Case-law trends clearly indicate that measurable improvements in algorithms—speed, accuracy, memory usage, or scalability—can serve as objective indicators of non-obviousness, especially when clearly documented and compared with prior art.
RELATED Blog
comments