IP Library › Granted Patent US 12,725,010
Granted Patent B2
US 12,725,010 · App. 19/181,522 · Granted Sep 1, 2026

Optimization of deep learning algorithms for large digital data processing using evolutionary neural networks

Inventor: Mohammadreza Nehzati (Berlin, DE)
Assignee: VMC MAR COM INC.
G06N3/0464G06N3/044G16H50/20
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,725,010
App. No.
19/181,522
Granted
Sep 1, 2026
Kind
B2
Abstract

Embodiments of the present disclosure relate to neural networks for processing large digital datasets. Neural networks comprise both a convolutional neural network (CNN) and a recurrent neural network (RNN). The neural networks are optimized by applying genetic algorithms. Embedding vectors are processed by both the CNN and the RNN to produce a merged output.

Claims (70)

1 . A method comprising:

initializing a first plurality of neural networks;

training the first plurality of neural networks on training data,

wherein the training data comprises artificial first electronic health records (EHRs) generated by combining elements of patient EHRs, and

wherein the first EHRs comprise patient demographics, medical diagnoses, medication information, laboratory test results, and medical imaging;

determining performance metrics, each of the performance metrics corresponding to a respective neural network of the first plurality of neural networks,

wherein each of the performance metrics is based on accuracy of the respective neural network;

selecting a subset of the first plurality of neural networks based on the performance metrics;

selecting a first neural network from the subset and a second neural network from the subset,

wherein the first neural network comprises a first convolutional neural network (CNN) and a first long short-term memory (LSTM) network, and

wherein the second neural network comprises a second CNN and a second LSTM network;

forming a third neural network, wherein forming the third neural network comprises (a) selecting weights of the first CNN, selecting weights of the second CNN, and forming a third CNN of the third neural network at least in part from the selected weights of the first CNN and the second CNN and (b) selecting weights of the first LSTM network, selecting weights of the second LSTM network, and forming a third LSTM network of the third neural network at least in part from the selected weights of the first LSTM network and the second LSTM network;

providing, to the third neural network, a plurality of embedding vectors generated from second EHRs;

processing, by the third CNN, a first subset of the embedding vectors, thereby producing a first output of the third CNN, wherein the third LSTM network does not process the first output of the third CNN;

processing, by the third CNN, a second subset of the embedding vectors, thereby producing a second output of the third CNN;

processing, by the third LSTM network, the second output of the third CNN, thereby producing an output of the third LSTM network;

combining the first output of the third CNN with the output of the third LSTM network, thereby producing a merged output;

classifying the second EHRs based on the merged output;

receiving a medical diagnosis for a patient, the patient corresponding to the second EHRs; and

generating a disease-progression prediction for the patient based on the medical diagnosis and the classification of the second EHRs.

2 . The method of claim 1 , further comprising:

determining that a convergence criterion has been satisfied based at least in part on an accuracy of the classification of the second EHRs.

3 . The method of claim 1 , further comprising applying random variations to weights of the third neural network.

4 . The method of claim 1 , wherein each of the first CNN, the second CNN, and the third CNN is a respective two-dimensional CNN.

5 . The method of claim 1 , wherein the first neural network and the second neural network are randomly selected from the subset.

6 . A system comprising:

a processor; and

memory storing instructions that, when executed by the processor, cause the processor to:

initialize a first plurality of neural networks;

train the first plurality of neural networks on training data,

wherein the training data comprises artificial first electronic health records (EHRs) generated by combining elements of patient EHRs, and

wherein the first EHRs comprise patient demographics, medical diagnoses, medication information, laboratory test results, and medical imaging;

determine performance metrics, each of the performance metrics corresponding to a respective neural network of the first plurality of neural networks,

wherein each of the performance metrics is based on accuracy of the respective neural network;

select a subset of the first plurality of neural networks based on the performance metrics;

select a first neural network from the subset and a second neural network from the subset,

wherein the first neural network comprises a first convolutional neural network (CNN) and a first long short-term memory (LSTM) network, and

wherein the second neural network comprises a second CNN and a second LSTM network;

form a third neural network, wherein forming the third neural network comprises (a) selecting weights of the first CNN, selecting weights of the second CNN, and forming a third CNN of the third neural network at least in part from the selected weights of the first CNN and the second CNN and (b) selecting weights of the first LSTM network, selecting weights of the second LSTM network, and forming a third LSTM network of the third neural network at least in part from the selected weights of the first LSTM network and the second LSTM network;

provide, to the third neural network, a plurality of embedding vectors generated from second EHRs;

process, by the third CNN, a first subset of the embedding vectors, thereby producing a first output of the third CNN, wherein the third LSTM network does not process the first output of the third CNN;

process, by the third CNN, a second subset of the embedding vectors, thereby producing a second output of the third CNN;

process, by the third LSTM network, the second output of the third CNN, thereby producing an output of the third LSTM network;

combine the first output of the third CNN with the output of the third LSTM network, thereby producing a merged output;

classify the second EHRs based on the merged output;

receive a medical diagnosis for a patient, the patient corresponding to the second EHRs; and

generate a disease-progression prediction for the patient based on the medical diagnosis and the classification of the second EHRs.

7 . A non-transitory computer readable medium having instructions thereon, the instructions, when executed by a computer, causing the computer to perform operations comprising:

initializing a first plurality of neural networks;

training the first plurality of neural networks on training data,

wherein the training data comprises artificial first electronic health records (EHRs) generated by combining elements of patient EHRs, and

wherein the first EHRs comprise patient demographics, medical diagnoses, medication information, laboratory test results, and medical imaging;

determining performance metrics, each of the performance metrics corresponding to a respective neural network of the first plurality of neural networks,

wherein each of the performance metrics is based on accuracy of the respective neural network;

selecting a subset of the first plurality of neural networks based on the performance metrics;

selecting a first neural network from the subset and a second neural network from the subset,

wherein the first neural network comprises a first convolutional neural network (CNN) and a first long short-term memory (LSTM) network, and

wherein the second neural network comprises a second CNN and a second LSTM network;

forming a third neural network, wherein forming the third neural network comprises (a) selecting weights of the first CNN, selecting weights of the second CNN, and forming a third CNN of the third neural network at least in part from the selected weights of the first CNN and the second CNN and (b) selecting weights of the first LSTM network, selecting weights of the second LSTM network, and forming a third LSTM network of the third neural network at least in part from the selected weights of the first LSTM network and the second LSTM network;

providing, to the third neural network, a plurality of embedding vectors generated from second EHRs;

processing, by the third CNN, a first subset of the embedding vectors, thereby producing a first output of the third CNN, wherein the third LSTM network does not process the first output of the third CNN;

processing, by the third CNN, a second subset of the embedding vectors, thereby producing a second output of the third CNN;

processing, by the third LSTM network, the second output of the third CNN, thereby producing an output of the third LSTM network;

combining the first output of the third CNN with the output of the third LSTM network, thereby producing a merged output;

classifying the second EHRs based on the merged output;

receiving a medical diagnosis for a patient, the patient corresponding to the second EHRs; and

generating a disease-progression prediction for the patient based on the medical diagnosis and the classification of the second EHRs.

8 . The method of claim 1 , wherein the merged output is based on a maximum of the first output of the third CNN and the output of the third LSTM network.

9 . The system of claim 6 , wherein the merged output is based on a maximum of the first output of the third CNN and the output of the third LSTM network.

10 . The medium of claim 7 , wherein the merged output is based on a maximum of the first output of the third CNN and the output of the third LSTM network.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Apr 17, 2025
From: NEHZATI, MOHAMMADREZA
To: VMC MAR COM INC.
Reel/Frame 070868/0522 →
Continuity (2)
Provisional Application 63675138 · Jul 24, 2024
Related Publication 20260030487A1 · Jan 29, 2026
References Cited (47)
US 9264483B2 · Hammond · 2016 [cited by applicant]
US 10015064B2 · Li et al. · 2018 [cited by applicant]
US 10657447B1 · McDonnell · 2020 [cited by examiner]
US 20190180186A1 · Liang · 2019 [cited by examiner]
US 20200380339A1 · Branson et al. · 2020 [cited by applicant]
US 20210209512A1 · Gaddam · 2021 [cited by examiner]
US 20230290502A1 · Zhi · 2023 [cited by examiner]
WO 2023114519A1 · 2023 [cited by applicant]
Kostrzewa, Daniel, Michal Ciszynski, and Robert Brzeski. “Evolvable hybrid ensembles for musical genre classification.” Proceedings of the Genetic and Evolutionary Computation Conference Companion. 2022. (Year: 2022). [cited by examiner]
Chiba, Zouhair, et al. “Automatic building of a powerful IDS for the cloud based on deep neural network by using a novel combination of simulated annealing algorithm and improved self-adaptive genetic algorithm.” Intern… [cited by examiner]
Choi, Keunwoo, et al. “Convolutional recurrent neural networks for music classification.” 2017 IEEE International conference on acoustics, speech and signal processing (ICASSP). IEEE, 2017. (Year: 2017). [cited by examiner]
Hao, Yixue, et al. “Recurrent convolutional neural network based multimodal disease risk prediction.” Future Generation Computer Systems 92 (2019): 76-83. (Year: 2019). [cited by examiner]
Gong, Jingtao, and Benjawan Srisura. “Heart Sound Detection Based on Bidirectional Multilayer Recurrent Convolutional Neural Network.” Proceedings of the 2022 4th International Conference on Software Engineering and Dev… [cited by examiner]
Xue, Ning, et al. “Evolving deep CNN-LSTMs for inventory time series prediction.” 2019 IEEE Congress on Evolutionary Computation (CEC). IEEE, 2019. (Year: 2019). [cited by examiner]
Yilmaz, Doga, et al. “Image Classification Using CNN-LSTM Hybrid Model With Skip Connections.” (Year: 2022). [cited by examiner]
Fu, Yingwei, et al. “Multi model-based distillation for sound event detection.” IEICE Transactions on Information and Systems 102.10 (2019): 2055-2058. (Year: 2019). [cited by examiner]
Deshpande, Pallavi, et al. “Alzheimer disease progression forecasting: Empowering models through hybrid of cnn and lstm with pso op-timization.” 2024 International Conference on Emerging Smart Computing and Informatics … [cited by examiner]
Saleh, Hager, et al. “Genetic algorithm-based hybrid deep learning model for explainable Alzheimer's disease prediction using temporal multimodal cognitive data.” International Journal of Data Science and Analytics 20.2… [cited by examiner]
Baniasadi et al., “Power Quality Disturbances Classification Using Identity Feature Vector and Support Vector Machine”, Journal of Soft Computing and Information Technology, 2020, pp. 151-164, vol. 9(2), English Abstrac… [cited by applicant]
Bhumika, P.S.S.S. et al., “A Review Paper on Algorithms Used for Text Classification”, International Journal of Application or Innovation in Engineering & Management, Mar. 2013, pp. 90-99, vol. 3(2). [cited by applicant]
Chen, C. et al., “Mitigating Backdoor Attacks in LSTM-Based Text Classification Systems by Backdoor Keyword Identification”, Neurocomputing, Mar. 15, 2021, pp. 1-13, vol. 452. 253-262. [cited by applicant]
Chen, Y.H. et al., “A hybrid text classification method based on K-congener-nearest-neighbors and hypersphere support vector machine”, 2013 International Conference on Information Technology and Applications, IEEE, Nov.… [cited by applicant]
Chen, J. et al., “Feature selection for text classification with Naïve Bayes”, Expert Systems with Applications, 2009, pp. 5432-5435, vol. 36. [cited by applicant]
Padurariu et al., “Dealing with Data Imbalance in Text Classification”, 23rd International Conference on Knowledge-Based and Intelligent Information & Engineering Systems, Procedia Computer Science, 2019, pp. 736-745, v… [cited by applicant]
Fang et al., “Automated text classification”, Advanced Engineering Informatics, Mar. 2, 2020, pp. 1-8, vol. 44. [cited by applicant]
Goel et al., “A Novel Feature Selection and Extraction Technique for Classification”, 14th International Conference on Frontiers in Handwriting Recognition, IEEE, 2014, pp. 104-109. [cited by applicant]
Golestanifar, et al., “Determination of Mental States from Texts Using Evolutionary Imperialist Competitive Algorithm and Convolution Neural Networks”, Journal of Soft Computing and Information Technology, 2021, pp. 13-… [cited by applicant]
Jang et al., “Sequential targeting: A continual learning approach for data imbalance in text classification”, Expert Systems With Applications, Apr. 21, 2021, pp. 1-9, vol. 179. [cited by applicant]
Kim et al., “Some Effective Techniques for Naive Bayes Text Classification”, IEEE Transactions on Knowledge and Data Engineering, Nov. 2006, pp. 1457-1466, vol. 18(11). [cited by applicant]
Li, Y., et al., “Imbalanced text sentiment classification using universal and domain-specific knowledge”, Knowledge-Based Systems, Jul. 2018, pp. 1-15, vol. 160. [cited by applicant]
Li, Z., et al., “End-to-End Adversarial Memory Network for Cross-domain Sentiment Classification”, Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17), Aug. 2017, pp. 223… [cited by applicant]
Luo et al., “Efficient English text classification using selected Machine Learning Techniques”, Alexandria Engineering Journal, Feb. 21, 2021, pp. 3401-3409, vol. 60. [cited by applicant]
Pop, I., “An approach of the Naive Bayes classifier for the document classification”, General Mathematics, 2006, pp. 135-138, vol. 14(4). [cited by applicant]
Singh et al., “Analysis of Vector Space Model in Information Retrieval”, Proceedings published by International Journal of Computer Applications, 2012, pp. 14-18. [cited by applicant]
Sun et al., “On strategies for imbalanced text classifification using SVM: A comparative study”, Decision Support Systems, 2009, pp. 191-201, vol. 48(1). [cited by applicant]
Tarekegn et al., “A review of methods for imbalance d multi-lab el classification”, Pattern Recognition, 2021, pp. 1-12, vol. 118. [cited by applicant]
Thabtah et al., “Data imbalance in classification: Experimental evaluation” Information Sciences, 2020, pp. 429-441, vol. 513. [cited by applicant]
Ting et al., “Is Naïve Bayes a Good Classifier for Document Classification?”, International Journal of Software Engineering and Its Applications, 2011, pp. 37-46, vol. 5(3). [cited by applicant]
Tsatsaronis et al., “A Generalized Vector Space Model for Text Retrieval Based on Semantic Relatedness”, Proceedings of the EACL 2009 Student Research Workshop, Apr. 2, 2009, pp. 70-78. [cited by applicant]
Xiao et al., “History-based attention in Seq2Seq model for multi-label text classification”, Knowledge-Based Systems, 2021, pp. 1-11, vol. 224. [cited by applicant]
Goldberg, “Genetic Algorithms in Search, Optimization, and Machine Learning”, Jan. 1989, Addison-Wesley Publishing Company, Inc., Cover, Forward, Preface, Contents, pp. 1-130. [cited by applicant]
Goldberg, “Genetic Algorithms in Search, Optimization, and Machine Learning”, Jan. 1989, Addison-Wesley Publishing Company, Inc., pp. 131-260. [cited by applicant]
Goldberg, “Genetic Algorithms in Search, Optimization, and Machine Learning”, Jan. 1989, Addison-Wesley Publishing Company, Inc., pp. 261-412. [cited by applicant]
Schmidhuber, “Deep learning in neural networks: An overview”, Neural Networks, 2015, pp. 85-114, vol. 61. [cited by applicant]
Xie et al., “Genetic CNN”, 2017 IEEE International Conference on Computer Vision (ICCV), 2017, pp. 1379-1388. [cited by applicant]
Nehzati, “Optimization of deep learning algorithms for large digital data processing usingnetworks”, Memories—Materials, Devices, Circuits and Systems, 2025, pp. 1-7. [cited by applicant]
European Patent Office; Extended European Search Report issued in corresponding EP 25187938.3, Nov. 26, 2025, pp. 1-9. [cited by applicant]