Low-cost device diagnoses diseases using deep learning.
problem Limited healthcare access and variable disease properties.
method Deep Neural Networks (DNNs) and Convolutional Neural Networks (CNNs) on a Raspberry Pi Zero.
result Achieves 90% accuracy for symptom-based diagnoses and 91% for visual skin diseases.
Paper develops an AI system to improve power system control.
problem Insufficient effectiveness of existing power system control.
method Combines deep learning and game theory for power system control.
result Improves power system control to normal steady-state or post-emergency conditions.
Paper tackles efficient allocation of multiple devices to users for AutoML services.
problem Allocating multiple devices to multiple users for AutoML services efficiently.
method Develops a multi-device, multi-tenant algorithm for GP-EI, achieving near-linear speedup.
result Achieves near-linear speedup when users are many more than devices.
Deep learning identifies unknown IoT devices in network traffic.
problem Unauthorized IoT devices pose security risks in BYOD environments.
method Deep learning applied to network traffic images for IoT device identification.
result Over 99% accuracy in identifying 10 IoT devices and non-white-listed devices.
Paper proposes low-rank gradient approximation to save memory for deep neural network training.
problem Memory limitation on mobile devices for deep neural network training.
method Approximating gradient matrices using low-rank parameterization.
result Reduces training memory by about 33.0% for Adam optimization and 4.5% relative lower word error rate on ASR personalization task.
Optimized CNNs for AMC on edge devices reduce complexity without sacrificing accuracy.
problem Developing efficient DL models for AMC on resource-constrained edge devices.
method Pruning, quantization, and knowledge distillation techniques applied to CNNs.
result Optimized models maintain or improve AMC accuracy with reduced complexity.
Single-Path NAS designs efficient ConvNets for mobile devices in hours.
problem Designing efficient ConvNets for mobile devices under latency constraints.
method Single-Path NAS, a differentiable method that reduces search cost and inference time.
result Achieves state-of-the-art accuracy on ImageNet with 79ms inference latency.
Improved speaker verification system detects replay attacks.
problem Detecting replay attacks in speaker verification systems.
method Multi-task learning of noise classes (replay noise, recording environment, recording device noises) and spoofing detection.
result Performance improved by 30% on ASVspoof2017 datasets.
Mobile training improves speech recognition for users with unique speech characteristics.
problem Limited generalization of speaker-independent speech recognition models for users with very different speech characteristics.
method Securely training personalized end-to-end speech recognition models on mobile devices, splitting gradient computation to reduce memory usage.
result On-device personalization achieved 58.1% relative word error rate reduction compared to 63.7% in a server environment, with 18.7% performance degradation.
This paper uses ML to identify prey handling in seals.
problem Automatically classify prey handling activity in seals for monitoring.
method Developed and compared three ML algorithms: Input Delay Neural Networks, Support Vector Machines, and Echo State Networks.
result Echo State Networks outperformed other algorithms in terms of accuracy and F1score.
FastDeepIoT optimizes neural network execution time on mobile devices.
problem Excessive neural network execution time on low-end devices.
method Automatically learns execution time model and optimizes network configurations.
result Significant reduction in execution time and energy consumption.
PruneFL reduces FL training time on edge devices by pruning model size.
problem Limited computation and communication resources on edge devices in FL.
method Adaptive and distributed parameter pruning during FL process.
result Pruned model converges to similar accuracy as original model with reduced training time.
User-specific KWS system learns new keywords on-device.
problem Out-of-vocabulary problem in traditional KWS systems.
method Query-by-example enrollment and testing, using phonetic posteriors and FST.
result Promising performance on two keywords, preserving simplicity.
New RL method optimizes power and accuracy for activity recognition.
problem Balancing power consumption and accuracy in wearable devices for activity recognition.
method Reinforcement Learning with multiple feedback sources for feature selection.
result Achieved good trade-off between power consumption and accuracy.
FedRule uses graph neural networks to recommend rules for smart homes without centralizing data.
problem Manual rule setup for smart devices is inefficient and privacy-compromising.
method FedRule constructs user-specific graphs for rule recommendation, using federated learning to protect privacy.
result FedRule achieves comparable performance to centralized methods and outperforms others.
Study improves voice disorder detection system robust to channel effects.
problem Voice signals are sensitive to recording devices.
method Bidirectional LSTM network with domain adversarial training (DAT).
result Increased PR-AUC from 0.8448 to 0.9455 (and 0.9522 with labels).
Method learns gestures from touch devices, outperforming state of the art.
problem Automatic gesture recognition on touch devices with multi-user variability.
method Dynamic sampling, convolutional model, recurrent vs. convolutional features.
result Outperforms state of the art on MMG dataset.
WEST compresses word embeddings and softmax layers for memory efficiency.
problem Memory constraints in large vocabulary models.
method WEST encodes words with sequences of sub-units, improving compression without performance loss.
result WEST achieves significant compression without sacrificing performance.
New method recovers compressed crack images for automatic segmentation.
problem Recover crack images from compressed data for SHM systems.
method Generative model-based CS method for crack images.
result Generative model effectively captures crack features for segmentation.
Evolutionary method reduces CNN complexity for mobile devices.
problem Efficiently compressing CNNs for mobile devices.
method Evolutionary algorithm to identify and remove redundant convolution filters.
result Generates an extremely compact CNN with improved compression and speed-up ratios.
A fall is an abnormal activity that occurs rarely, so it is hard to collect real data for falls. It is, therefore, difficult to use supervised learning methods to automatically detect falls. Another challenge in using machine learning methods to automatically detect falls is the choice of engineered features. In this p…
Develops panoramic gastroscopy for automatic polyp detection.
problem Missed diagnosis of gastric polyps during endoscopy.
method Panoramic reconstruction method and end-to-end multi-object detection.
result Average error of panorama less than 2 mm, polyp detection accuracy 95%, recall rate 99%.
A new method for automatically aligning and clustering time series data.
problem Challenges in aligning and clustering time series data, especially without a template signal.
method TROUT (Temporal Registration using Optimal Unitary Transformations) method based on a novel dissimilarity measure.
result TROUT outperforms competitors in clustering time series data.
A wearable device-based sleep stage classifier using feature learning and RNNs.
problem Automatic sleep stage classification using wearable devices.
method Multi-level feature learning framework and RNN classifier with BLSTM.
result The algorithm achieves high precision, recall, and F1 scores in both resting and comprehensive groups.
Paper presents CeNN quantization for efficient CPS applications.
problem Efficient processing for CPS applications, especially in telemedicine and ADAS.
method Incremental quantization of CeNNs with various strategies.
result Achieved up to 7.8x speedup with no performance loss.
Automates design of lightweight neural networks for image classification.
problem Designing efficient neural networks for edge devices with limited computational resources.
method Uses the Mesh Adaptive Direct Search (MADS) algorithm to optimize network architecture.
result Achieves comparable performance to standard methods with fewer design trials.
Paper proposes efficient pruning method for neural networks.
problem Compressing deep neural networks for resource-constrained devices.
method Adaptive sparsity loss for budget-aware optimization during training.
result Demonstrated effectiveness on various architectures and datasets.
Galen algorithm compresses neural networks for specific hardware with reduced latency.
problem Finding optimal compression policies for neural networks on specific hardware.
method Reinforcement learning using pruning and quantization to optimize inference latency.
result Compressed ResNet18 for ARM processor reduced inference latency by 80%.
HASP enhances mobile security against ASR by adding imperceptible noise.
problem Security leakage from ASR on mobile devices.
method Adversarial noise to perturb WER and enhance security.
result HASP reduces WER by 84.55% and speeds up data processing.
This work automates CNN model compression for mobile devices.
problem Deploying trained CNNs to mobile devices requires balancing speed, memory, and accuracy.
method Reinforcement learning with architecture search and knowledge distillation.
result An automated model compression algorithm improves the trade-off between speed, memory, and accuracy.
Iterative AutoML improves ASR model compression by 5x without WER degradation.
problem Challenges in achieving high compression levels without degrading ASR performance.
method Iterative AutoML-based Low Rank Factorization (LRF) approach.
result Achieved over 5x compression without WER degradation.
GeneCAI optimizes DNN compression hyper-parameters for mobile devices.
problem Efficient deployment of complex DNNs on resource-limited devices.
method GeneCAI uses genetic algorithm to learn optimal hyper-parameters.
result GeneCAI finds models with better accuracy-complexity trade-off.
GShard enables scaling of large neural networks with automatic sharding and lightweight APIs.
problem Scaling neural networks to handle vast training data and compute efficiently.
method GShard uses lightweight annotation APIs and XLA compiler extensions for parallel computation.
result GShard successfully trained a 600 billion parameter model on 2048 TPUs in 4 days.
Paper proposes FTT-NAS to create fault-tolerant CNNs for edge devices.
problem Faults in edge devices affect deep learning applications.
method Formalized fault models, implemented FTT-NAS, incorporated FTT.
result Discovered CNNs outperform baseline architectures with fault tolerance.
Julia programs and ML models are compiled and offloaded to Google Cloud TPUs.
problem Efficiently compiling and executing machine learning models on Google Cloud TPUs.
method Developed a method to compile and offload Julia programs and ML models to Google Cloud TPUs using the Google XLA compiler.
result Successfully fused a VGG19 model into a single TPU executable, achieving 0.23s evaluation time on a batch of 100 images.
AutoQ automatically optimizes quantization for CNNs, reducing latency and energy.
problem Efficiently quantizing CNN weights for low-power mobile devices.
method Hierarchical-DRL for kernel-wise quantization bitwidth selection.
result Reduces inference latency and energy consumption by 54.06% and 50.69% respectively.
Paper uses UKS to improve BLE RSSI for proximity inference in mobile phone apps.
problem Improper BLE RSSI for accurate proximity inference during pandemics.
method Single-dimensional Unscented Kalman Smoother (UKS) with Gaussian process observation transforms.
result UKS outperforms traditional methods in predicting infection risk from BLE RSSI.
Tiny Eats GRU detects eating episodes on a microcontroller.
problem Automatic dietary monitoring on low-power devices.
method Shallow gated recurrent unit (GRU) architecture on Arm Cortex M0+.
result Tiny Eats GRU achieves 95.15% accuracy with 4% memory usage and 6 ms latency.
MemNet optimizes neural architectures for memory efficiency.
problem Memory constraints in mobile devices limit the use of large neural networks.
method Augment-trim learning with memory consumption ranking score.
result MemNet finds architectures with 24.17% less memory usage compared to state-of-the-art methods.
FedZKT enables resource-constrained devices to participate in federated learning with heterogeneous models.
problem Inequality in resource allocation hinders participation from resource-constrained devices in federated learning.
method Zero-shot knowledge transfer through a server-assigned distillation process.
result FedZKT effectively transfers knowledge across heterogeneous on-device models without requiring comparable local training efforts.
This paper proposes an automatic neural network compression method.
problem Reducing resource requirements for deep neural networks on resource-constrained devices.
method Jointly prunes and quantizes neural networks without manual hyper-parameter tuning.
result Significant reduction in model size with minimal accuracy loss.
Proposes a channel pruning method using attention statistics for deep networks.
problem Manual setting of compression ratios in each layer for deep neural networks.
method Channel pruning based on attention statistics with automatic selection of compression ratio.
result Improved performance in terms of accuracy and computational costs compared to conventional methods.
On-device federated learning updates edge models by exchanging trained results.
problem Limited training data at edge devices due to model drift.
method OS-ELM for sequential training and autoencoder for anomaly detection, combined with federated learning.
result The proposed approach produces a merged model as accurately as traditional methods with lower costs.
SGQuant reduces GNN memory usage without significant accuracy loss.
problem High memory consumption in GNNs limits their applicability on memory-constrained devices.
method Proposes a specialized GNN quantization scheme (SGQuant) with a quantization algorithm, fine-tuning scheme, and multi-granularity strategy.
result SGQuant reduces GNN memory footprint from 4.25x to 31.9x with minimal accuracy loss.
RILOD enables edge devices to learn new object classes efficiently.
problem Edge devices need to learn new object classes without losing old class detection.
method RILOD uses a novel incremental learning algorithm that trains end-to-end for one-stage deep object detection models.
result RILOD can learn to detect a new object class in just a few minutes.
A method for trust evaluation of devices in human-device coexistence systems.
problem Efficient trust evaluation of devices in systems with diverse physical and social attributes.
method Canonical correlation analysis-enhanced hypergraph self-supervised learning (HSLCCA).
result The proposed HSLCCA method significantly outperforms baseline algorithms in identifying trusted devices.
Study proposes a new approval policy for ML-based medical devices to prevent gradual performance degradation.
problem Gradual deterioration in machine learning model performance over time in medical devices.
method Formulated an automatic algorithmic change protocol (aACP) as an online hypothesis testing problem, considering both error-rate guarantees and non-guaranteed policies.
result Controlled the rate of gradual deterioration (biocreep) in machine learning models without significantly impacting approval of beneficial modifications.
Two approaches scale up DNN optimization for diverse edge devices.
problem Optimizing DNNs for edge devices with varying performance requirements.
method Reuse performance predictors on proxy devices and build scalable predictors.
result Optimized DNN designs for many different edge devices without lengthy optimization.