• Next-generation risk-aware XAI for mission-critical applications

• Advances trust and transparency in high-stakes decision making

REDMOND, Wash., May 19, 2025 (GLOBE NEWSWIRE) -- Pattern Computer®, Inc. (“Pattern” or “the Company”), the global leader in Pattern Discovery, today announced the addition of RC/XAI™ (Risk and Confidence Explainable AI), to its Pattern Discovery Engine™ (PDE). The immediate effect of this technical achievement will be to bring known and demonstrated confidence and risk values to decisions for customers using the company’s new PatternDE™ online platform.

Users in industries such as equity market trading are expected to be among the first to benefit from this unique capacity, adding new assurance on a per – trade basis.

This release introduces advanced risk-aware model convergence, conformal prediction, and enhanced explainability, setting a new standard for trustworthy AI in mission-critical sectors such as finance, medical diagnostics, and industrial process optimization. These new features derive from additions to the Iceland™ PDE component.

Innovative Features in Iceland/SR™ 3.0

Version 3.0 incorporates several groundbreaking enhancements, including the Janus-Delphi Loss Framework™, risk-aware and instance-level metrics, and a refined conformal gating regime. Key features include:

• Janus-Delphi Loss Framework – A new evolution of the Janus Loss system, integrating per-instance scalable overconfidence penalties and localized risk-sensitive metrics. This approach enhances posterior probability estimation, ensuring models are finely calibrated to their operating context. The Janus-Delphi framework supports critical domain requirements for informed-consent by decision makers relying on the models' predictions, including per-instance certainty scores and real-time risk stratification.
  
• Risk-Aware and Risk-Transparent Model Generation – Iceland/SR 3.0 introduces risk-aware symbolic regression, allowing models to self-calibrate through dynamically adjusted confidence metrics. Moreover, the symbolic models (mathematical equations interpretable directly) explicitly self-optimize during training to maximize the risk-surface they expose for analysis, allowing end-users informed consent when deciding whether or not to base decisions on their predictions. This transparency not only facilitates high-confidence predictions while minimizing overfitting, but also supports compliance with emerging regulatory frameworks such as the SEC’s proposed AI risk management rules, the EU’s AI Act, and other global standards aimed at increasing accountability in automated decision-making.
  
• Abstention Metrics for Risk and Confidence – Newly added abstention gating mechanisms provide instance-level coverage, decision gating scores, and confidence intervals, enabling precise quantification of model risk, lost opportunity expectations when gates are trusted, and predicted abstention rates in real-world deployments. This allows users to make data-driven decisions with quantifiable confidence.
  
• Information-Theoretic and Conformity Metrics – Expanded support for information-theoretic measures to optimize model complexity and interpretability, ensuring each prediction is grounded in academic and industry standard, statistically rigorous post hoc evaluation and selection for best generalization on as-yet-unseen data.
  

Redefining Explainability in AI

With its highly optimized multi-threaded architecture and robust symbolic inference capabilities, RC/XAI represents a significant move forward for organizations seeking both cutting-edge performance and transparent model behavior.

Mark R. Anderson, Pattern Founding Chair and CEO, commented, "What started as comparatively simple equations-as-predictors in our earlier Gilford Island™ system has become a full decision-support trust architecture in Iceland/SR™. Explainable AI has reached a new frontier: one where it holds itself accountable. This is not merely about understanding a model’s decision path but about calibrating it with enough precision to allow for truly informed consent in high-stakes environments. These are models that don’t just tell you how they arrive at a decision, but also how self-confident they are in each specific instance. This is a breakthrough in a new mathematics of “trust-aware AI.”

Pattern has developed the new technology in conjunction with well-known Wall Street firms. Anderson added, "With this new technology, we’re delivering more than just mathematical models—we’re setting a new benchmark for transparency and trust in explainable AI. This isn’t just about better algorithms, it’s about empowering organizations to make informed decisions that answer to their risk tolerance policies. Our mission has always been to discover previously unavailable patterns in data, and with RC/XAI and this new version of PatternDE, we’re now giving our customers and partners the clarity they need to act with confidence in even the most complex, high-stakes environments.”

Availability

Iceland/SR 3.0 and RC/XAI are now available in Pattern’s PatternDE™ platform.

About Pattern
Pattern Computer, Inc. uses its Pattern Discovery Engine™ to solve the most important and intractable problems in business and medicine. These proprietary mathematical techniques in advanced AI can find complex patterns in very high-dimensional data that have eluded detection by much larger systems. As the Company applies its computational platform to the challenging fields of drug discovery and diagnostics, it is also making major Pattern Discoveries for partners in other sectors, including extended biotech, materials science, aerospace manufacturing quality control, veterinary medicine, air traffic operations, energy services, finance, market trading and more. See www.patterncomputer.com.

CONTACT: Laura Guerrant-Oiye (808) 960-2642 – laura@patterncomputer.com

The foregoing contains statements about Pattern Computer’s future that are not statements of historical fact. These statements are “forward looking statements” for purposes of applicable securities laws and are based on current information and/or management’s good faith belief as to future events. The words “believe,” “expect,” “anticipate,” “project,” “should,” “could,” “will,” and similar expressions signify forward-looking statements. Forward-looking statements should not be read as a guarantee of future performance. By their nature, forward-looking statements involve inherent risk and uncertainties, which change over time, and actual performance could differ materially from that anticipated by any forward-looking statements. Pattern Computer undertakes no obligation to update or revise any forward-looking statement.

Copyright © 2025 Pattern Computer Inc. All Rights Reserved. Pattern Computer, Inc., Pattern Discovery Engine, PatternBio, TrueXAI, and ProSpectral are trademarks of Pattern Computer Inc. or its subsidiaries. Other trademarks may be trademarks of their respective owners.