Invention Grant
- Patent Title: Systems and methods to identify neural network brittleness based on sample data and seed generation
-
Application No.: US18383946Application Date: 2023-10-26
-
Publication No.: US12277455B2Publication Date: 2025-04-15
- Inventor: Austin Walters , Vincent Pham , Galen Rafferty , Anh Truong , Mark Watson , Jeremy Goodsitt
- Applicant: CAPITAL ONE SERVICES, LLC
- Applicant Address: US VA McLean
- Assignee: CAPITAL ONE SERVICES, LLC
- Current Assignee: CAPITAL ONE SERVICES, LLC
- Current Assignee Address: US VA McLean
- Agency: Foley & Lardner LLP
- Main IPC: G06N20/00
- IPC: G06N20/00 ; G06F8/71 ; G06F9/54 ; G06F11/3604 ; G06F11/362 ; G06F16/22 ; G06F16/242 ; G06F16/2455 ; G06F16/248 ; G06F16/25 ; G06F16/28 ; G06F16/335 ; G06F16/903 ; G06F16/9032 ; G06F16/9038 ; G06F16/906 ; G06F16/93 ; G06F17/15 ; G06F17/16 ; G06F17/18 ; G06F18/20 ; G06F18/21 ; G06F18/2115 ; G06F18/214 ; G06F18/22 ; G06F18/23 ; G06F18/24 ; G06F18/2411 ; G06F18/2415 ; G06F18/40 ; G06F21/55 ; G06F21/60 ; G06F21/62 ; G06F30/20 ; G06F40/117 ; G06F40/166 ; G06F40/20 ; G06N3/04 ; G06N3/044 ; G06N3/045 ; G06N3/06 ; G06N3/08 ; G06N3/088 ; G06N3/094 ; G06N5/00 ; G06N5/02 ; G06N5/04 ; G06N7/00 ; G06N7/01 ; G06Q10/04 ; G06T7/194 ; G06T7/246 ; G06T7/254 ; G06T11/00 ; G06V10/70 ; G06V10/98 ; G06V30/194 ; G06V30/196 ; H04L9/40 ; H04L67/00 ; H04L67/306 ; H04N21/234 ; H04N21/81

Abstract:
Systems and methods for determining neural network brittleness are disclosed. For example, the system may include one or more memory units storing instructions and one or more processors configured to execute the instructions to perform operations. The operations may include receiving a modeling request comprising a preliminary model and a dataset. The operations may include determining a preliminary brittleness score of the preliminary model. The operations may include identifying a reference model and determining a reference brittleness score of the reference model. The operations may include comparing the preliminary brittleness score to the reference brittleness score and generating a preferred model based on the comparison. The operations may include providing the preferred model.
Public/Granted literature
- US20240054029A1 SYSTEMS AND METHODS TO IDENTIFY NEURAL NETWORK BRITTLENESS BASED ON SAMPLE DATA AND SEED GENERATION Public/Granted day:2024-02-15
Information query