EvoMBN: Evolving Multi-Branch Networks on Myocardial Infarction Diagnosis Using 12-Lead Electrocardiograms

Multi-branch Networks (MBNs) have been successfully applied to myocardial infarction (MI) diagnosis using 12-lead electrocardiograms. However, most existing MBNs share a fixed architecture. The absence of architecture optimization has become a significant obstacle to a more accurate diagnosis for these MBNs. In this paper, an evolving neural network named EvoMBN is proposed for MI diagnosis. It utilizes a genetic algorithm (GA) to automatically learn the optimal MBN architectures. A novel fixed-length encoding method is proposed to represent each architecture. In addition, the crossover, mutation, selection, and fitness evaluation of the GA are defined to ensure the architecture can be optimized through evolutional iterations. A novel Lead Squeeze and Excitation (LSE) block is designed to summarize features from all the branch networks. It consists of a fully-connected layer and an LSE mechanism that assigns weights to different leads. Five-fold inter-patient cross validation experiments on MI detection and localization are performed using the PTB diagnostic database. Moreover, the model architecture learned from the PTB database is transferred to the PTB-XL database without any changes. Compared with existing studies, our EvoMBN shows superior generalization and the efficiency of its flexible architecture is suitable for auxiliary MI diagnosis in real-world.

Design Optimization of B-series Marine Propeller using NSGA-II, Iterative and Gekko Algorithm

The design of a propeller plays a significant role in naval architecture. Optimization of various design factors is the primary concern for effective and efficient propulsion. This study investigates the optimization of the B-series marine propellers using three different methods, i.e. (i) a non-linear constrained single-objective optimization approach using the Non-Dominated Sorting Genetic Algorithm (NSGA-II), (ii) a python package for dynamic optimization based optimization software ‘Gekko’, (iii) an iterative approach and results were compared with each other. Efficiency is considered as the single objective function whereas three constraints are imposed: cavitation, thrust and strength. Analogous characteristics have been found in the comparison of results from all three methods. Comparing the various factors, this study suggests that, Gekko can be used as the optimization algorithm.

Automatic Composition of Encoding Scheme and Search Operators in System Architecture Optimization

The purpose of this paper is to propose a new method for the automatic composition of both encoding schemes and search operators for system architecture optimization. The method leverages prior work that identified a set of six patterns that appear often in system architecture decision problems (down-selecting, combining, assigning, partitioning, permuting, and connecting). First, the user models the architecture space as a directed graph, where nodes are decisions belonging to one of the aforementioned patterns, and edges are dependencies between decisions that affect architecture enumeration (e.g., the outcome of decision 1 affects the number of alternatives available for decision 2). Then, based on a library of encoding scheme and operator fragments that are appropriate for each pattern, an algorithm automatically composes an encoding scheme and corresponding search operators by traversing the graph. The method is demonstrated in two case studies: a study integrating three architectural decisions for constructing a portfolio of earth observing satellite missions, and a study integrating eight architectural decisions for designing a guidance navigation and control system. Results suggest that this method has comparable search performance to hand-crafted formulations from experts. Furthermore, the proposed method drastically reducing the need for practitioners to write new code.

A Compression-Compilation Framework for On-mobile Real-time BERT Applications

Transformer-based deep learning models have increasingly demonstrated high accuracy on many natural language processing (NLP) tasks. In this paper, we propose a compression-compilation co-design framework that can guarantee the identified model meets both resource and real-time specifications of mobile devices. Our framework applies a compiler-aware neural architecture optimization method (CANAO), which can generate the optimal compressed model that balances both accuracy and latency. We are able to achieve up to 7.8x speedup compared with TensorFlow-Lite with only minor accuracy loss. We present two types of BERT applications on mobile devices: Question Answering (QA) and Text Generation. Both can be executed in real-time with latency as low as 45ms. Videos for demonstrating the framework can be found on https://www.youtube.com/watch?v=_WIRvK_2PZI