Andac Demir

Title: Andac Demir: Innovator in Neural Network Architecture

Introduction

Andac Demir is a prominent inventor based in Boston, MA, known for his contributions to the field of artificial intelligence and neural networks. He has developed innovative systems that enhance the construction and efficiency of neural network architectures.

Latest Patents

One of his notable patents is titled "Automated construction of neural network architecture with Bayesian graph exploration." This patent describes a system designed for the automated construction of an artificial neural network architecture. The system includes a set of interfaces and data links that are configured to receive and send signals, which encompass datasets of training, validation, and testing data. It also incorporates random number factors in multi-dimensional signals, which are associated with task labels and nuisance variations. The system features memory banks that store reconfigurable deep neural network (DNN) blocks, hyperparameters, and trainable variables. A processor is integrated to execute Bayesian graph exploration, optimizing the DNN blocks by pruning redundant links and modifying hyperparameters. This innovation allows for nuisance-robust variational Bayesian inference, making it adaptable to new datasets in semi-supervised settings.

Career Highlights

Andac Demir is currently employed at Mitsubishi Electric Research Laboratories, Inc., where he continues to push the boundaries of technology in neural networks. His work has significant implications for the development of more efficient and effective artificial intelligence systems.

Collaborations

Throughout his career, Andac has collaborated with esteemed colleagues such as Toshiaki Koike-Akino and Ye Wang, contributing to advancements in the field of neural networks.

Conclusion

Andac Demir's innovative work in neural network architecture exemplifies the potential of artificial intelligence to transform technology. His contributions are paving the way for future advancements in the field.