Search

Editor’s notes: This article discusses current directions in computation-aware deep learning and describes two new challenges in the IoT era: 1) data-independent deployment of learning and 2) communication-aware distributed inference. It presents new directions to alleviate these challenges. — Deming Chen, University of Illinois at Urbana-Champaign

In this paper, we first highlight three major challenges to large-scale adoption of deep learning at the edge: (i) Hardware-constrained IoT devices, (ii) Data security and privacy in the IoT era, and (iii) Lack of network-aware deep learning algorithms for distributed inference across multiple IoT devices. We then provide a unified view targeting three research directions that naturally emerge from the above challenges: (1) Federated learning for training deep networks, (2) Data-independent deployment of learning algorithms, and (3) Communication-aware distributed inference. We believe that the above research directions need a network-centric approach to enable the edge intelligence and, therefore, fully exploit the true potential of IoT.

This paper presents a hardware prototype and a framework for a new communication-aware model compression for distributed on-device inference. Our approach relies on Knowledge Distillation (KD) and achieves orders of magnitude compression ratios on a large pre-trained teacher model. The distributed hardware prototype consists of multiple student models deployed on Raspberry-Pi 3 nodes that run Wide ResNet and VGG models on the CIFAR10 dataset for real-time image classification. We observe significant reductions in memory footprint (50×), energy consumption (14×), latency (33×) and an increase in performance (12×) without any significant accuracy loss compared to the initial teacher model. This is an important step towards deploying deep learning models for IoT applications.

In this paper, we present a new approach to interpret deep learning models. By coupling mutual information with network science, we explore how information flows through feedforward networks. We show that efficiently approximating mutual information allows us to create an information measure that quantifies how much information flows between any two neurons of a deep learning model. To that end, we propose NIF, Neural Information Flow, a technique for codifying information flow that exposes deep learning model internals and provides feature attributions., Comment: ICASSP 2020 (shorter version appeared at AAAI-19 Workshop on Network Interpretability for Deep Learning)

Model compression has emerged as an important area of research for deploying deep learning models on Internet-of-Things (IoT). However, for extremely memory-constrained scenarios, even the compressed models cannot fit within the memory of a single device and, as a result, must be distributed across multiple devices. This leads to a distributed inference paradigm in which memory and communication costs represent a major bottleneck. Yet, existing model compression techniques are not communication-aware. Therefore, we propose Network of Neural Networks (NoNN), a new distributed IoT learning paradigm that compresses a large pretrained 'teacher' deep network into several disjoint and highly-compressed 'student' modules, without loss of accuracy. Moreover, we propose a network science-based knowledge partitioning algorithm for the teacher model, and then train individual students on the resulting disjoint partitions. Extensive experimentation on five image classification datasets, for user-defined memory/performance budgets, show that NoNN achieves higher accuracy than several baselines and similar accuracy as the teacher model, while using minimal communication among students. Finally, as a case study, we deploy the proposed model for CIFAR-10 dataset on edge devices and demonstrate significant improvements in memory footprint (up to 24x), performance (up to 12x), and energy per node (up to 14x) compared to the large teacher model. We further show that for distributed inference on multiple edge devices, our proposed NoNN model results in up to 33x reduction in total latency w.r.t. a state-of-the-art model compression baseline., Comment: This preprint is for personal use only. The official article will appear as part of the ESWEEK-TECS special issue and will be presented in the International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS), 2019

Real world networks constructed from raw data are often characterized by complex community structures. Existing dimensionality reduction techniques, however, do not take such characteristics into account. This is especially important for problems with low number of samples where the curse of dimensionality is particularly significant. Therefore, in this paper, we propose FeatureNet, a novel community-based dimensionality reduction framework targeting small sample problems. To this end, we propose a new method to directly construct a network from high-dimensional raw data while explicitly revealing its hidden community structure; these communities are then used to learn low-dimensional features using a representation learning framework. We show the effectiveness of our approach on eight datasets covering application areas as diverse as handwritten digits, biology, physical sciences, NLP, and computational sustainability. Extensive experiments on the above datasets (with sizes mostly between 100 and 1500 samples) demonstrate that FeatureNet significantly outperforms (i.e., up to 40% improvement in classification accuracy) ten well-known dimensionality reduction methods like PCA, Kernel PCA, Isomap, SNE, t-SNE, etc.

Forecasting short-term river flowrate is an important problem for flood planning/preparation and hydropower management. In this paper, we propose K-Hop Learning, a novel feature extraction method, which utilizes network information to learn upcoming river flow peaks 24-hours in advance. We then use these learned features to forecast flowrate 24-hours ahead at various locations in the Ohio River Basin, USA. Our results clearly demonstrate that the predictions using K-Hop Learning are far superior to several traditional timeseries-based, machine learning models such as SVM Regression. Further, in contrast to these existing timeseries-based models, the upcoming river flow peaks are predicted very accurately by our approach. Overall, using K-Hop Learning, we significantly reduce the prediction RMSE by 57--82% depending on the river size.