Transfer learning on edge is challenging due to on-device limited resources. Existing work addresses this issue by training a subset of parameters or adding model patches. Developed with inference in mind, Inverted Residual Blocks (IRBs) split a convolutional layer into depthwise and pointwise convolutions, leading to more stacking layers, e.g., convolution, normalization, and activation layers. Though they are efficient for inference, IRBs require that additional activation maps are stored in memory for training weights for convolution layers and scales for normalization layers. As a result, their high memory cost prohibits training IRBs on resource-limited edge devices, and making them unsuitable in the context of transfer learning. To address this issue, we present MobileTL, a memory and computationally efficient on-device transfer learning method for models built with IRBs. MobileTL trains the shifts for internal normalization layers to avoid storing activation maps for the backward pass. Also, MobileTL approximates the backward computation of the activation layer (e.g., Hard-Swish and ReLU6) as a signed function which enables storing a binary mask instead of activation maps for the backward pass. MobileTL fine-tunes a few top blocks (close to output) rather than propagating the gradient through the whole network to reduce the computation cost. Our method reduces memory usage by 46% and 53% for MobileNetV2 and V3 IRBs, respectively. For MobileNetV3, we observe a 36% reduction in floating-point operations (FLOPs) when fine-tuning 5 blocks, while only incurring a 0.6% accuracy reduction on CIFAR10. Extensive experiments on multiple datasets demonstrate that our method is Pareto-optimal (best accuracy under given hardware constraints) compared to prior work in transfer learning for edge devices.
more »
« less
This content will become publicly available on July 1, 2024
TIPS: Topologically Important Path Sampling for Anytime Neural Networks
Anytime neural networks (AnytimeNNs) are a promising solution to adaptively adjust the model complexity at runtime under various hardware resource constraints. However, the manually-designed AnytimeNNs are biased by designers' prior experience and thus provide sub-optimal solutions. To address the limitations of existing hand-crafted approaches, we first model the training process of AnytimeNNs as a discrete-time Markov chain (DTMC) and use it to identify the paths that contribute the most to the training of AnytimeNNs. Based on this new DTMC-based analysis, we further propose TIPS, a framework to automatically design AnytimeNNs under various hardware constraints. Our experimental results show that TIPS can improve the convergence rate and test accuracy of AnytimeNNs. Compared to the existing AnytimeNNs approaches, TIPS improves the accuracy by 2%-6.6% on multiple datasets and achieves SOTA accuracy-FLOPs tradeoffs.
more »
« less
- Award ID(s):
- 2007284
- NSF-PAR ID:
- 10468130
- Publisher / Repository:
- International Conference on Machine Learning (ICML)
- Date Published:
- Subject(s) / Keyword(s):
- ["Deep Learning, Anytime Neural Networks, Discrete-Time Markov Chain"]
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Neural architecture search (NAS) is a promising technique to design efficient and high-performance deep neural networks (DNNs). As the performance requirements of ML applications grow continuously, the hardware accelerators start playing a central role in DNN design. This trend makes NAS even more complicated and time-consuming for most real applications. This paper proposes FLASH, a very fast NAS methodology that co-optimizes the DNN accuracy and performance on a real hardware platform. As the main theoretical contribution, we first propose the NN-Degree, an analytical metric to quantify the topological characteristics of DNNs with skip connections (e.g., DenseNets, ResNets, Wide-ResNets, and MobileNets). The newly proposed NN-Degree allows us to do training-free NAS within one second and build an accuracy predictor by training as few as 25 samples out of a vast search space with more than 63 billion configurations. Second, by performing inference on the target hardware, we fine-tune and validate our analytical models to estimate the latency, area, and energy consumption of various DNN architectures while executing standard ML datasets. Third, we construct a hierarchical algorithm based on simplicial homology global optimization (SHGO) to optimize the model-architecture co-design process, while considering the area, latency, and energy consumption of the target hardware. We demonstrate that, compared to the state-of-the-art NAS approaches, our proposed hierarchical SHGO-based algorithm enables more than four orders of magnitude speedup (specifically, the execution time of the proposed algorithm is about 0.1 seconds). Finally, our experimental evaluations show that FLASH is easily transferable to different hardware architectures, thus enabling us to do NAS on a Raspberry Pi-3B processor in less than 3 seconds.more » « less
-
Large-scale deep neural networks (DNNs) are both compute and memory intensive. As the size of DNNs continues to grow, it is critical to improve the energy efficiency and performance while maintaining accuracy. For DNNs, the model size is an important factor affecting performance, scalability and energy efficiency. Weight pruning achieves good compression ratios but suffers from three drawbacks: 1) the irregular network structure after pruning, which affects performance and throughput; 2) the increased training complexity; and 3) the lack of rigirous guarantee of compression ratio and inference accuracy. To overcome these limitations, this paper proposes CirCNN, a principled approach to represent weights and process neural networks using block-circulant matrices. CirCNN utilizes the Fast Fourier Transform (FFT)-based fast multiplication, simultaneously reducing the computational complexity (both in inference and training) from O(n2) to O(n log n) and the storage complexity from O(n2) to O(n), with negligible accuracy loss. Compared to other approaches, CirCNN is distinct due to its mathematical rigor: the DNNs based on CirCNN can converge to the same "effectiveness" as DNNs without compression. We propose the CirCNN architecture, a universal DNN inference engine that can be implemented in various hardware/software platforms with configurable network architecture (e.g., layer type, size, scales, etc.). In CirCNN architecture: 1) Due to the recursive property, FFT can be used as the key computing kernel, which ensures universal and small-footprint implementations. 2) The compressed but regular network structure avoids the pitfalls of the network pruning and facilitates high performance and throughput with highly pipelined and parallel design. To demonstrate the performance and energy efficiency, we test CirCNN in FPGA, ASIC and embedded processors. Our results show that CirCNN architecture achieves very high energy efficiency and performance with a small hardware footprint. Based on the FPGA implementation and ASIC synthesis results, CirCNN achieves 6 - 102X energy efficiency improvements compared with the best state-of-the-art results.more » « less
-
he pervasive operation of customer drones, or small-scale unmanned aerial vehicles (UAVs), has raised serious concerns about their privacy threats to the public. In recent years, privacy invasion events caused by customer drones have been frequently reported. Given such a fact, timely detection of invading drones has become an emerging task. Existing solutions using active radar, video or acoustic sensors are usually too costly (especially for individuals) or exhibit various constraints (e.g., requiring visual line of sight). Recent research on drone detection with passive RF signals provides an opportunity for low-cost deployment of drone detectors on commodity wireless devices. However, the state of the arts in this direction rely on line-of-sight (LOS) RF signals, which makes them only work under very constrained conditions. The support of more common scenarios, i.e., non-line-of-sight (NLOS), is still missing for low-cost solutions. In this paper, we propose a novel detection system for privacy invasion caused by customer drone. Our system is featured with accurate NLOS detection with low-cost hardware (under $50). By exploring and validating the relationship between drone motions and RF signal under the NLOS condition, we find that RF signatures of drones are somewhat “amplified” by multipaths in NLOS. Based on this observation, we design a two-step solution which first classifies received RSS measurements into LOS and NLOS categories; deep learning is then used to extract the signatures and ultimately detect the drones. Our experimental results show that LOS and NLOS signals can be identified at accuracy rates of 98.4% and 96% respectively. Our drone detection rate for NLOS condition is above 97% with a system implemented using Raspberry PI 3 B+.more » « less
-
null (Ed.)Residential homes constitute roughly one-fourth of the total energy usage worldwide. Providing appliance-level energy breakdown has been shown to induce positive behavioral changes that can reduce energy consumption by 15%. Existing approaches for energy breakdown either require hardware installation in every target home or demand a large set of energy sensor data available for model training. However, very few homes in the world have installed sub-meters (sensors measuring individual appliance energy); and the cost of retrofitting a home with extensive sub-metering eats into the funds available for energy saving retrofits. As a result, strategically deploying sensing hardware to maximize the reconstruction accuracy of sub-metered readings in non-instrumented homes while minimizing deployment costs becomes necessary and promising. In this work, we develop an active learning solution based on low-rank tensor completion for energy breakdown. We propose to actively deploy energy sensors to appliances from selected homes, with a goal to improve the prediction accuracy of the completed tensor with minimum sensor deployment cost. We empirically evaluate our approach on the largest public energy dataset collected in Austin, Texas, USA, from 2013 to 2017. The results show that our approach gives better performance with fixed number of sensors installed, when compared to the state-of-the-art, which is also proven by our theoretical analysis.more » « less
