No abstract

Most digital cameras use sensors coated with a Color Filter Array (CFA) to capture channel components at every pixel location, resulting in a mosaic image that does not contain pixel values in all channels. Current research on reconstructing these missing channels, also known as demosaicing, introduces many artifacts, such as zipper effect and false color. Many deep learning demosaicing techniques outperform other classical techniques in reducing the impact of artifacts. However, most of these models tend to be over-parametrized. Consequently, edge implementation of the state-of-the-art deep learning-based demosaicing algorithms on low-end edge devices is a major challenge. We provide an exhaustive search of deep neural network architectures and obtain a Pareto front of Color Peak Signal to Noise Ratio (CPSNR) as the performance criterion versus the number of parameters as the model complexity that outperforms the state-of-the-art. Architectures on the Pareto front can then be used to choose the best architecture for a variety of resource constraints. Simple architecture search methods such as exhaustive search and grid search requires some conditions of the loss function to converge to the optimum. We clarify these conditions in a brief theoretical study.

No abstract

No abstract

Tiny machine learning (tinyML) has emerged during the past few years aiming to deploy machine learning models to embedded AI processors with highly constrained memory and computation capacity. Low precision quantization is an important model compression technique that can greatly reduce both memory consumption and computation cost of model inference. In this study, we focus on post-training quantization (PTQ) algorithms that quantize a model to low-bit (less than 8-bit) precision with only a small set of calibration data and benchmark them on different tinyML use cases.To achieve a fair comparison, we build a simulated quantization framework to investigate recent PTQ algorithms. Furthermore, we break down those algorithms into essential components and reassembled a generic PTQ pipeline. With ablation study on different alternatives of components in the pipeline, we reveal key design choices when performing low precision quantization. We hope this work could provide useful data points and shed lights on the future research of low precision quantization.

No abstract

Recurrent neural networks (RNN) such as long-short-term memory (LSTM) networks are essential in a multitude of daily live tasks such as speech, language, video, and multimodal learning. The shift from cloud to edge computation intensifies the need to contain the growth of RNN parameters. Current research on RNN shows that despite the performance obtained on convolutional neural networks (CNN), keeping a good performance in compressed RNNs is still a challenge. Most of the literature on compression focuses on CNNs using matrix product (MPO) operator tensor trains. However, matrix product state (MPS) tensor trains have more attractive features than MPOs, in terms of storage reduction and computing time at inference. We show that MPS tensor trains should be at the forefront of LSTM network compression through a theoretical analysis and practical experiments on NLP tasks.Since the beginning of the last decade, computational resources have taken a primary role in the advancing of artificial intelligence. With the advent of new architectures for neural networks, starting with AlexNet and convolutional neural networks, computational demand has accelerated to the point of doubling itself each hundred days [Amodei et al., 2019]. In fact from AlexNet to AlexZero, computing demand has increased 3000 times. This implies major challenges for resources such as memory. Neural network model size is a bottleneck. The use of memory and computer power for training neural nets is enormous. For example, AlexNet takes 1.5 weeks to train on ImageNet on one Nvidia TitanX GPU [Han et al., 2015]. Although this might seems like a long time, we have to consider that following the current trend in neural network modeling, a more sophisticated architecture, such as ResNet152, might take at least ten times longer to train. Model size is indeed a major challenge: larger model implies larger sizes, which implies higher energy consumption. In a recent tweet debate, Elliot Turner, the CEO and co-founder of Hologram AI, wrote that it costs $ Preprint. Under review.

No abstract