Tesseract Systems LLC v. Mitre Corp

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Tesseract Systems LLC (NM)
  • Defendant: The Mitre Corporation (DE)
  • Plaintiff’s Counsel: Rabicoff Law LLC
• Case Identification: 7:25-cv-00166, W.D. Tex., 04/16/2025
• Venue Allegations: Venue is asserted based on the defendant having an established place of business in the Western District of Texas.
• Core Dispute: Plaintiff alleges that Defendant’s unspecified products, which relate to neural networks, infringe a patent directed to a specific neural network architecture.
• Technical Context: The technology concerns "highway networks," a specialized deep neural network architecture designed to ease the training of networks with very large numbers of layers.
• Key Procedural History: The complaint does not mention any prior litigation, inter partes review (IPR) proceedings, or licensing history related to the patent-in-suit. The claim of willfulness is based on knowledge derived from the filing of this complaint.

Case Timeline

Date	Event
2016-05-02	Priority Date for U.S. Patent No. 10,984,320
2021-04-20	Issue Date for U.S. Patent No. 10,984,320
2025-04-16	Complaint Filed

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 10,984,320 - "Highly trainable neural network configuration"

• Patent Identification: U.S. Patent No. 10,984,320, issued April 20, 2021.

The Invention Explained

• Problem Addressed: The patent’s background section identifies the difficulty of training deep neural networks, noting that as layers are added, optimization becomes "considerably more difficult" than for shallower networks (’320 Patent, col. 1:37-40).
• The Patented Solution: The patent discloses a novel neural network architecture, termed a "highway network," that facilitates the training of very deep networks (’320 Patent, col. 2:53-54). This is achieved through a "learned gating mechanism for regulating information flow," which allows information to pass across many layers "without attenuation" (’320 Patent, col. 2:35-38). As illustrated in Figure 1, individual neurons are configured with "transform" and "carry" gates that learn to dynamically blend a transformed version of the input signal with a carried-forward, non-transformed component of the input signal, creating an "information highway" that bypasses some transformations (’320 Patent, col. 4:3-23).
• Technical Importance: This architecture enables the training of extremely deep networks (e.g., up to 100 layers or more), which was previously hampered by optimization problems, thereby allowing researchers to study the impact of network depth on complex problems with fewer restrictions (’320 Patent, col. 2:53-60, col. 13:3-7).

Key Claims at a Glance

• The complaint does not specify which claims are asserted, instead referring to "Exemplary '320 Patent Claims" identified in a separate, unprovided exhibit (Compl. ¶11). Independent claim 1 is a representative method claim.
• The essential elements of independent claim 1 include:
  • Receiving an electrical input signal at a neuron.
  • Applying a first non-linear transform to the input signal to produce a "plain signal."
  • Applying a second non-linear transform at a first gate to produce a "transform signal."
  • Applying a third non-linear transform at a second gate to produce a "carry signal."
  • Calculating a weighted sum of a "non-transformed first component of the input signal" and the plain signal, which is accomplished by multiplying the plain signal by the transform signal and, separately, multiplying the non-transformed input component by the carry signal, and then adding the two products.
• The complaint does not explicitly reserve the right to assert dependent claims but refers generally to infringement of "one or more claims" (Compl. ¶11).

III. The Accused Instrumentality

Product Identification

• The complaint does not name any specific accused products, methods, or services, referring only to "the Defendant products identified in the charts incorporated into this Count" as "Exemplary Defendant Products" (Compl. ¶11). These charts were provided in an exhibit that is not part of the public complaint document.

Functionality and Market Context

• The complaint does not provide sufficient detail for analysis of the accused instrumentality's functionality or market context. It makes only the conclusory allegation that the "Exemplary Defendant Products practice the technology claimed by the '320 Patent" (Compl. ¶16).

IV. Analysis of Infringement Allegations

The complaint does not contain claim charts or a narrative infringement theory in the body of the document, instead incorporating them by reference from an unattached "Exhibit 2" (Compl. ¶16-17). The complaint alleges that the defendant directly infringes by making, using, selling, and importing the accused products, and also by its employees' internal use and testing of those products (Compl. ¶11-12). Without access to Exhibit 2, a detailed analysis of the infringement allegations is not possible.

No probative visual evidence provided in complaint.

• Identified Points of Contention:
  • Factual Question: A central factual dispute will be whether the accused products, once identified, implement the specific gating architecture of the asserted claims. The key question is what evidence exists that the accused products use distinct "transform" and "carry" signals to calculate a weighted sum of a transformed signal (the "plain signal") and a "non-transformed" component of the input signal.
  • Scope Questions: The infringement analysis raises the question of whether the functionality of the accused products meets the specific claim limitations. For instance, does an accused system that applies any form of processing to all data paths still have a "non-transformed first component of the input signal" as required by claim 1?

V. Key Claim Terms for Construction

• The Term: "non-transformed first component of the input signal" (Claim 1)

• Context and Importance: This term is central to the invention's "highway" concept, defining the bypass path that allows information to flow without transformation. The outcome of the infringement analysis may depend on whether the accused product's data path can be characterized as containing a "non-transformed" component.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The patent does not provide an explicit definition for "non-transformed." A party could argue that in the context of the claim, "non-transformed" simply means "not subjected to the first non-linear transform," potentially allowing for other minor processing like buffering or scaling.
  • Evidence for a Narrower Interpretation: The specification's primary mathematical formula for the invention is y=H(x, WH)·T(x, Wt)+x·(1-T(x, Wt)) ('320 Patent, col. 11:38-40). The use of x (the input) on the right side of the addition suggests that the original, unmodified input itself is being carried forward. This supports a narrower construction requiring a data path for a component of the original input signal that is free of substantive transformation.

• The Term: "plain signal" (Claim 1)

• Context and Importance: This term appears to be the patentee's own lexicography. Practitioners may focus on this term because its definition is tied to the output of the "first non-linear transform," and its relationship to the other signals is critical for understanding the claimed neuron operation.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: As a patentee-defined term, its scope could be argued to encompass any intermediate signal resulting from the primary non-linear function (H) of the neuron, regardless of the specific function used.
  • Evidence for a Narrower Interpretation: Claim 1 explicitly defines the term by its creation: "applying a first non-linear transform to the input signal at the neuron to produce a plain signal" ('320 Patent, col. 17:12-14). This suggests its meaning is not independent and must be construed strictly as the direct output of the H function, as described in the specification (y=H(x, WH)) ('320 Patent, col. 11:21).

VI. Other Allegations

• Indirect Infringement: The complaint pleads induced infringement, alleging that Defendant distributes "product literature and website materials inducing end users and others to use its products in the customary and intended manner that infringes" the '320 Patent (Compl. ¶14).
• Willful Infringement: The allegation of willfulness is based on post-suit conduct. The complaint asserts that the service of the complaint itself provides "actual knowledge of infringement" and that any subsequent infringing activities by the Defendant are therefore willful (Compl. ¶13-14). No facts are alleged to support pre-suit knowledge.

VII. Analyst’s Conclusion: Key Questions for the Case

1. Evidentiary Sufficiency: A threshold issue, given the complaint's reliance on an unprovided exhibit, will be whether the plaintiff's infringement contentions, once fully disclosed, meet the plausibility standard for pleading under federal rules.
2. Architectural Correspondence: The central technical question will be whether the accused neural network products implement the specific three-transform, two-gate "highway" architecture recited in the asserted claims. Specifically, do they calculate an output by blending a transformed signal with a "non-transformed" component of the input, as controlled by a carry gate?
3. Definitional Scope: The case may turn on claim construction, particularly the meaning of the term "non-transformed first component of the input signal". The court's interpretation of this term will determine the scope of the bypass or "highway" path required to prove infringement.