Convolutional neural network detects early seizures with low power microcontroller.
problem Early detection of seizures for patients with severe epilepsy.
method Energy-efficient convolutional neural network designed for implantable microcontrollers.
result Outperforms other detectors with high sensitivity and low false detection rate.
This paper enables deep network inference on microcontrollers with improved accuracy and reduced memory usage.
problem Deploying deep networks on resource-constrained edge-devices with low memory and computational constraints.
method Mixed low-bitwidth compression, rule-based iterative procedure for bit precision determination, quantization-aware retraining, and integer-only model conversion.
result Improved Top1 accuracy of 68% on a 2MB FLASH memory STM32H7 microcontroller, 8% higher than 8-bit implementations.
Paper predicts embryo implantation probability from IVF time-lapse imaging.
problem Manual embryo selection in IVF has low success rates.
method Data-driven system trained on time-lapse imaging videos.
result Algorithm improves positive and negative predictive values.
Researchers save memory on MCUs by reordering neural network operators.
problem Memory constraints on microcontrollers for neural network inference.
method Operator reordering to save memory, orthogonal to other compression methods.
result Reduced memory footprint of a CNN to deploy on an MCU with 512KB SRAM.
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.
FANN-on-MCU enables efficient neural network inference on IoT devices.
problem Energy-efficient neural network inference on resource-constrained IoT devices.
method Open-source toolkit for ARM Cortex-M and RISC-V microcontrollers.
result Efficient neural network execution with low latency and power consumption.
This work optimizes DNN inference for energy-harvesting devices by compressing and selectively executing neural network exits.
problem Inference delays and energy inefficiency in energy-harvesting devices.
method Developed a power trace-aware and exit-guided network compression algorithm for multi-exit neural networks.
result Superior accuracy and reduced latency compared to state-of-the-art techniques.
Improved deep learning model deployment on tiny MCUs with mixed-precision quantization.
problem Memory limitations prevent accurate deployment of DNN models on tiny MCUs.
method Automated mixed-precision quantization using Reinforcement Learning for MCU constraints.
result Mixed-precision models achieve high accuracy with uniform quantization policies.
Backdoors can be implanted in neural models of source code, and we detect and remove them.
problem Vulnerability of neural models to backdoors in source code.
method Defined and implemented various backdoor classes, adapted robust statistics algorithms, and detected poisoned data through spectral signatures.
result Demonstrated the ease of injecting and removing backdoors in neural models of source code.
Generative compression technique reduces neural network size and improves performance on microcontrollers.
problem Memory constraints on microcontrollers limit the size of neural networks, especially for 1x1 pointwise (PW) mixers.
method HYPER-TINYPW uses a shared micro-MLP to generate PW kernels from tiny per-layer codes, reducing memory usage.
result HYPER-TINYPW achieves comparable performance to larger models while being significantly smaller (225 kB vs 1.4 MB).
Research creates a machine learning model for predicting TAVI patient mortality.
problem Lack of robust risk scores for TAVI patients.
method Gradient boosting on decision trees, feature analysis and selection, model validation.
result Model outperforms existing risk scores with AUC of 0.83.
TCA algorithm saves up to 40% energy in IoT devices.
problem Energy consumption and user comfort trade-off in IoT devices.
method Incremental learning Temporal Clustering Algorithm (TCA) for IoT devices.
result Achieved up to 40% energy savings with high success rate.
Personalized deep learning reduces inappropriate shocks in VA detection.
problem High inappropriate shock rate in traditional VA detection methods.
method Personalized deep learning framework using CNN for real-time VA detection and collaborative inference.
result 6.6% reduction in inappropriate shock rate compared to traditional methods.
This paper introduces a novel graph-analytic approach for detecting anomalies in network flow data called GraphPrints. Building on foundational network-mining techniques, our method represents time slices of traffic as a graph, then counts graphlets -- small induced subgraphs that describe local topology. By performing…
Resource-efficient oblique trees reduce neural signal classification costs.
problem Implementing efficient neural signal classifiers on resource-constrained devices.
method Integrating model compression, probabilistic routing, and cost-aware learning.
result Significant reduction in model size and feature extraction cost compared to state-of-the-art models.
A new HL-SVR approach handles unequal sample sizes in SVR for engineering data modeling.
problem SVR assumes equal sample sizes, but unequal sizes are common in engineering.
method HL-SVR combines low-level SVR for larger samples and high-level SVR for smaller samples.
result HL-SVR produces more accurate predictions than conventional SVR.
This thesis optimizes neuromorphic systems by slowing down their dynamics, improving performance.
problem Timescale mismatch between analog neuromorphic circuits and real-time sensory inputs.
method Proposes and tests solutions to slow down the dynamics of spiking neural networks.
result Spiking neural networks on analog neuromorphic systems can achieve significant performance boosts.
New KWS neural networks improve accuracy and power efficiency.
problem Maximizing accuracy and power efficiency in KWS systems.
method Hardware Aware Training (HAT) for LMU-based neural networks.
result Achieved state-of-the-art accuracy and low parameter counts.
Unique 3-balls in S^4 can be non-isotopic.
problem Existence and uniqueness of spanning 3-balls in S^4.
method Introducing barbell diffeomorphisms, implantations, and twistings; 2-parameter calculus of embeddings; framed cobordism method.
result Non-uniqueness of spanning 3-balls in S^4.
Fine-tuning neural networks to guarantee performance on specific examples can also introduce incorrect inputs.
problem Ensuring reliable performance of neural networks on specific examples.
method Using SMT solvers to fine-tune ReLU neural networks to guarantee outcomes on a finite set of particular examples.
result Fine-tuning can introduce incorrect inputs that trigger unexpected performance.
This project was motivated by a dialysis study in northern Taiwan. Dialysis patients, after shunt implantation, may experience two types ("acute" or "non-acute") of shunt thrombosis, both of which may recur. We formulate the problem under the framework of recurrent events data in the presence of competing risks. In par…
Efficient classifier with uncertainty bounds for safety-critical applications.
problem Lack of uncertainty bounds in high-accuracy classifiers for safety-critical tasks.
method Nadaraya-Watson estimator with frequentist bounds.
result Competitive accuracy and uncertainty bounds at reduced computational cost.
SEFR is a fast, energy-efficient classifier for ultra-low power devices.
problem Running machine learning on battery-powered devices is challenging due to time and energy constraints.
method SEFR is an ultra-low power classifier with linear time complexity for training and testing.
result SEFR is 63 times faster and 70 times more energy efficient than state-of-the-art classifiers.
Enhances neural networks' robustness against adversarial samples without sacrificing clean sample generalization.
problem Limited generalization and time complexity of adversarial training.
method Feature Pyramid Decoder (FPD) framework that integrates denoising and image restoration modules into CNNs and constrains the Lipschitz constant.
result FPD-enhanced CNNs achieve sufficient robustness against general adversarial samples on various datasets.
Paper explores embedding methods for detecting pseudo-cliques in random graphs, showing limitations and potential.
problem Detecting planted pseudo-cliques in random dot product graphs.
method Adjacency Spectral Embedding (ASE) and Graph Encoder Embedding (GEE).
result These methods can localize pseudo-cliques with additional clean network data, but not without it.
DNI recovers missing brain data from corrupted recordings.
problem Corrupted neural recordings from multielectrode systems.
method Deep Neural Imputation framework using autoencoders.
result DNI recovers both time series and frequency content from corrupted data.
Improved cardiac arrhythmia detection in wearable devices with neural networks.
problem Resource constraints in low-power wearable devices for accurate arrhythmia detection.
method Adapted a convolutional-recurrent neural network to a low-power microcontroller, optimizing for precision and memory usage.
result Reduced F1 score from 0.8 to 0.784 in fixed-point precision, with a 195.6KB memory footprint and 33.98MOps/s throughput. Constructs pseudo-isotopies with Cerf diagrams and Hatcher-Wagoner invariants for barbell maps.
problem Understanding and constructing pseudo-isotopies for barbell maps.
method Constructs specific pseudo-isotopies with Cerf diagrams and Hatcher-Wagoner invariants.
result Every pseudo-isotopy with vanishing first Hatcher-Wagoner invariant can be isotoped to a composition of standard barbell pseudo-isotopies.
EAST compresses deep ConvNets for tiny memory nodes.
problem Memory constraints in tiny devices for deep ConvNets.
method Encoding-Aware Sparse Training (EAST) with adaptive group pruning and LZ4 weight encoding.
result EAST achieves deep memory compression with lower sparsity and higher accuracy.
Developing a Brain-Computer Interface~(BCI) for seizure prediction can help epileptic patients have a better quality of life. However, there are many difficulties and challenges in developing such a system as a real-life support for patients. Because of the nonstationary nature of EEG signals, normal and seizure patter…
TinyLSTMs reduces speech enhancement model size and latency for hearing aids.
problem Large RNNs limit practical deployment in hearing aid hardware.
method Model compression techniques (pruning, integer quantization, state update skipping) for RNN speech enhancement.
result Reduction in model size and operations by 11.9imes and 2.9imes, respectively, without perceptual degradation. Low-cost sensor fusion for organic substance classification.
problem Non-destructive classification and quality control of organic substances.
method Arduino Mega 2560 microcontroller with three sensors, supervised learning models, hyperparameter tuning, PCA/LDA, ensemble voting classifier.
result Test accuracies of 93-94% for identifying organic compounds.
This paper proposes IMU preintegrated features for efficient deep inertial odometry.
problem Efficient odometry from IMU data is challenging due to sensor imperfections and noise.
method Proposes IMU preintegrated features exploiting IMU motion model's manifold structure.
result Improves odometry performance and reduces computational burdens.
New method converts conventional ANNs to SNNs with minimal loss and efficiency.
problem Difficulty in training SNNs directly from conventional ANNs due to discreteness.
method Proposes a novel pipeline combining threshold balance and soft-reset mechanisms for efficient conversion.
result Achieves almost no accuracy loss with only 1/10 of typical SNN simulation time.
Study geodesic X-ray transform and streaking artifacts on simple surfaces or spaces of constant curvature.
problem Streaking artifacts in CT images due to metal regions.
method Geodesic X-ray transform on nontrapping compact Riemannian manifolds with strictly convex boundaries.
result Streaking artifacts result from conormal singularities along common tangent geodesics.
SIMCom detects HTs by analyzing communication patterns in SoCs.
problem Timely detection of Hardware Trojans in secure integrated circuits.
method Multi-parameter statistical modeling of communication channels in SoCs.
result SIMCom detects HTs with less than 1% area and power overhead.
FastGRNN improves RNN accuracy while drastically reducing model size.
problem Inaccurate training and inefficient prediction in RNNs.
method FastGRNN uses a residual connection and gate to achieve state-of-the-art accuracy with a much smaller model.
result FastGRNN achieves state-of-the-art accuracy with models up to 35x smaller than existing RNNs.
Review of efficient neural networks for TinyML on resource-constrained devices.
problem Resource constraints on ultra-low power MCUs for deep learning models.
method Model compression, quantization, low-rank factorization, model pruning, hardware acceleration, algorithm-architecture co-design.
result Optimized neural network architectures for minimal resource utilization on MCUs.
New BMI interface stabilizes user experience over long periods.
problem Inconsistent neural signal data leads to interface recalibration.
method Adversarial Domain Adaptation Network to match residual distributions.
result Adversarial Domain Adaptation Network outperforms other methods.
When the full stock of a new product is quickly sold in a few days or weeks, one has the impression that new technologies develop and conquer the market in a very easy way. This may be true for some new technologies, for example the cell phone, but not for others, like the blue-ray. Novelty, usefulness, advertising, pr…
Edge-based sensor network monitors natural hazards efficiently.
problem Fast decision making in natural hazard warning systems.
method Event-triggered micro-seismic sensors, co-detection, machine learning, convolutional neural networks.
result Real-time hazard monitoring with low power consumption.
Computer vision performances have been significantly improved in recent years by Convolutional Neural Networks(CNN). Currently, applications using CNN algorithms are deployed mainly on general purpose hardwares, such as CPUs, GPUs or FPGAs. However, power consumption, speed, accuracy, memory footprint, and die size sho…
Paper proposes a training framework for deploying DNNs on analog NVM crossbars, achieving significant efficiency gains.
problem Challenges in deploying deep learning models on microcontrollers due to limited compute and memory resources.
method Developed a training algorithm to eliminate tuning and propose unipolar-weighted matrices to reduce crossbar area and simplify hardware.
result Achieved up to 92.91% accuracy with 2-bit crossbars and up to 45% energy reduction.
Tall wheatgrass outperforms rye in energy and environmental metrics, marginally improving economic viability.
problem Finding sustainable alternatives for marginal agricultural areas.
method Economic assessment using profit margin and Life Cycle Assessment (LCA) for energy and environmental performance.
result Tall wheatgrass shows better environmental and energy performance with reduced inputs and lower energy consumption.