The hardware-software co-optimization of neural network architectures is a field of research that emerged with the advent of commercial neuromorphic chips, such as the IBM TrueNorth and Intel Loihi. ...Development of simulation and automated mapping software tools in tandem with the design of neuromorphic hardware, whilst taking into consideration the hardware constraints, will play an increasingly significant role in deployment of system-level applications. This paper illustrates the importance and benefits of co-design of convolutional neural networks (CNN) that are to be mapped onto neuromorphic hardware with a crossbar array of synapses. Toward this end, we first study which convolution techniques are more hardware friendly and propose different mapping techniques for different convolutions. We show that, for a seven-layered CNN, our proposed mapping technique can reduce the number of cores used by 4.9-13.8 times for crossbar sizes ranging from 128 × 256 to 1,024 × 1,024, and this can be compared to the toeplitz method of mapping. We next develop an iterative co-design process for the systematic design of more hardware-friendly CNNs whilst considering hardware constraints, such as core sizes. A python wrapper, developed for the mapping process, is also useful for validating hardware design and studies on traffic volume and energy consumption. Finally, a new neural network dubbed HFNet is proposed using the above co-design process; it achieves a classification accuracy of 71.3% on the IMAGENET dataset (comparable to the VGG-16) but uses 11 times less cores for neuromorphic hardware with core size of 1,024 × 1,024. We also modified the HFNet to fit onto different core sizes and report on the corresponding classification accuracies. Various aspects of the paper are patent pending.
IMPORTANCE: Deep learning (DL) networks require large data sets for training, which can be challenging to collect clinically. Generative models could be used to generate large numbers of synthetic ...optical coherence tomography (OCT) images to train such DL networks for glaucoma detection. OBJECTIVE: To assess whether generative models can synthesize circumpapillary optic nerve head OCT images of normal and glaucomatous eyes and determine the usability of synthetic images for training DL models for glaucoma detection. DESIGN, SETTING, AND PARTICIPANTS: Progressively growing generative adversarial network models were trained to generate circumpapillary OCT scans. Image gradeability and authenticity were evaluated on a clinical set of 100 real and 100 synthetic images by 2 clinical experts. DL networks for glaucoma detection were trained with real or synthetic images and evaluated on independent internal and external test data sets of 140 and 300 real images, respectively. MAIN OUTCOMES AND MEASURES: Evaluations of the clinical set between the experts were compared. Glaucoma detection performance of the DL networks was assessed using area under the curve (AUC) analysis. Class activation maps provided visualizations of the regions contributing to the respective classifications. RESULTS: A total of 990 normal and 862 glaucomatous eyes were analyzed. Evaluations of the clinical set were similar for gradeability (expert 1: 92.0%; expert 2: 93.0%) and authenticity (expert 1: 51.8%; expert 2: 51.3%). The best-performing DL network trained on synthetic images had AUC scores of 0.97 (95% CI, 0.95-0.99) on the internal test data set and 0.90 (95% CI, 0.87-0.93) on the external test data set, compared with AUCs of 0.96 (95% CI, 0.94-0.99) on the internal test data set and 0.84 (95% CI, 0.80-0.87) on the external test data set for the network trained with real images. An increase in the AUC for the synthetic DL network was observed with the use of larger synthetic data set sizes. Class activation maps showed that the regions of the synthetic images contributing to glaucoma detection were generally similar to that of real images. CONCLUSIONS AND RELEVANCE: DL networks trained with synthetic OCT images for glaucoma detection were comparable with networks trained with real images. These results suggest potential use of generative models in the training of DL networks and as a means of data sharing across institutions without patient information confidentiality issues.
The hardware-software co-optimization of neural network architectures is becoming a major stream of research especially due to the emergence of commercial neuromorphic chips such as the IBM Truenorth ...and Intel Loihi. Development of specific neural network architectures in tandem with the design of the neuromorphic hardware considering the hardware constraints will make a huge impact in the complete system level application. In this paper, we study various neural network architectures and propose one that is hardware-friendly for a neuromorphic hardware with crossbar array of synapses. Considering the hardware constraints, we demonstrate how one may design the neuromorphic hardware so as to maximize classification accuracy in the trained network architecture, while concurrently, we choose a neural network architecture so as to maximize utilization in the neuromorphic cores. We also proposed a framework for mapping a neural network onto a neuromorphic chip named as the Mapping and Debugging (MaD) framework. The MaD framework is designed to be generic in the sense that it is a Python wrapper which in principle can be integrated with any simulator tool for neuromorphic chips.
Neuromorphic systems or dedicated hardware for neuromorphic computing is getting popular with the advancement in research on different device materials for synapses, especially in crossbar ...architecture and also algorithms specific or compatible to neuromorphic hardware. Hence, an automated mapping of any deep neural network onto the neuromorphic chip with crossbar array of synapses and an efficient debugging framework is very essential. Here, mapping is defined as the deployment of a section of deep neural network layer onto a neuromorphic core and the generation of connection lists among population of neurons to specify the connectivity between various neuromorphic cores on the neuromorphic chip. Debugging is the verification of computations performed on the neuromorphic chip during inferencing. Together the framework becomes Mapping and Debugging (MaD) framework. MaD framework is quite general in usage as it is a Python wrapper which can be integrated with almost every simulator tools for neuromorphic chips. This paper illustrates the MaD framework in detail, considering some optimizations while mapping onto a single neuromorphic core. A classification task on MNIST and CIFAR-10 datasets are considered for test case implementation of MaD framework.