arXiv subjects

Tong Lin

Publications and source records attributed to Tong Lin.

46 records · Page 3Linked to original sources

arXiv2022

Unsupervised Domain Adaptation with Histogram-gated Image Translation for Delayered IC Image Analysis

Deep learning has achieved great success in the challenging circuit annotation task by employing Convolutional Neural Networks (CNN) for the segmentation of circuit structures. The deep learning approaches require a large amount of manually annotated training data to achieve a good performance, which could cause a degradation in performance if a deep learning model trained on a given dataset is applied to a different dataset. This is commonly known as the domain shift problem for circuit annotation, which stems from the possibly large variations in distribution across different image datasets. The different image datasets could be obtained from different devices or different layers within a single device. To address the domain shift problem, we propose Histogram-gated Image Translation (HGIT), an unsupervised domain adaptation framework which transforms images from a given source dataset to the domain of a target dataset, and utilize the transformed images for training a segmentation network. Specifically, our HGIT performs generative adversarial network (GAN)-based image translation and utilizes histogram statistics for data curation. Experiments were conducted on a single labeled source dataset adapted to three different target datasets (without labels for training) and the segmentation performance was evaluated for each target dataset. We have demonstrated that our method achieves the best performance compared to the reported domain adaptation techniques, and is also reasonably close to the fully supervised benchmark.

cs.CV↗

arXiv2021

Boosting Image Outpainting with Semantic Layout Prediction

The objective of image outpainting is to extend image current border and generate new regions based on known ones. Previous methods adopt generative adversarial networks (GANs) to synthesize realistic images. However, the lack of explicit semantic representation leads to blurry and abnormal image pixels when the outpainting areas are complex and with various objects. In this work, we decompose the outpainting task into two stages. Firstly, we train a GAN to extend regions in semantic segmentation domain instead of image domain. Secondly, another GAN model is trained to synthesize real images based on the extended semantic layouts. The first model focuses on low frequent context such as sizes, classes and other semantic cues while the second model focuses on high frequent context like color and texture. By this design, our approach can handle semantic clues more easily and hence works better in complex scenarios. We evaluate our framework on various datasets and make quantitative and qualitative analysis. Experiments demonstrate that our method generates reasonable extended semantic layouts and images, outperforming state-of-the-art models.

cs.CV↗

arXiv2021

Principal Gradient Direction and Confidence Reservoir Sampling for Continual Learning

Task-free online continual learning aims to alleviate catastrophic forgetting of the learner on a non-iid data stream. Experience Replay (ER) is a SOTA continual learning method, which is broadly used as the backbone algorithm for other replay-based methods. However, the training strategy of ER is too simple to take full advantage of replayed examples and its reservoir sampling strategy is also suboptimal. In this work, we propose a general proximal gradient framework so that ER can be viewed as a special case. We further propose two improvements accordingly: Principal Gradient Direction (PGD) and Confidence Reservoir Sampling (CRS). In Principal Gradient Direction, we optimize a target gradient that not only represents the major contribution of past gradients, but also retains the new knowledge of the current gradient. We then present Confidence Reservoir Sampling for maintaining a more informative memory buffer based on a margin-based metric that measures the value of stored examples. Experiments substantiate the effectiveness of both our improvements and our new algorithm consistently boosts the performance of MIR-replay, a SOTA ER-based method: our algorithm increases the average accuracy up to 7.9% and reduces forgetting up to 15.4% on four datasets.

cs.LG↗

arXiv2021

A New Adaptive Gradient Method with Gradient Decomposition

Adaptive gradient methods, especially Adam-type methods (such as Adam, AMSGrad, and AdaBound), have been proposed to speed up the training process with an element-wise scaling term on learning rates. However, they often generalize poorly compared with stochastic gradient descent (SGD) and its accelerated schemes such as SGD with momentum (SGDM). In this paper, we propose a new adaptive method called DecGD, which simultaneously achieves good generalization like SGDM and obtain rapid convergence like Adam-type methods. In particular, DecGD decomposes the current gradient into the product of two terms including a surrogate gradient and a loss based vector. Our method adjusts the learning rates adaptively according to the current loss based vector instead of the squared gradients used in Adam-type methods. The intuition for adaptive learning rates of DecGD is that a good optimizer, in general cases, needs to decrease the learning rates as the loss decreases, which is similar to the learning rates decay scheduling technique. Therefore, DecGD gets a rapid convergence in the early phases of training and controls the effective learning rates according to the loss based vectors which help lead to a better generalization. Convergence analysis is discussed in both convex and non-convex situations. Finally, empirical results on widely-used tasks and models demonstrate that DecGD shows better generalization performance than SGDM and rapid convergence like Adam-type methods.

cs.LG↗

arXiv2021

Intra-Model Collaborative Learning of Neural Networks

Recently, collaborative learning proposed by Song and Chai has achieved remarkable improvements in image classification tasks by simultaneously training multiple classifier heads. However, huge memory footprints required by such multi-head structures may hinder the training of large-capacity baseline models. The natural question is how to achieve collaborative learning within a single network without duplicating any modules. In this paper, we propose four ways of collaborative learning among different parts of a single network with negligible engineering efforts. To improve the robustness of the network, we leverage the consistency of the output layer and intermediate layers for training under the collaborative learning framework. Besides, the similarity of intermediate representation and convolution kernel is also introduced to reduce the reduce redundant in a neural network. Compared to the method of Song and Chai, our framework further considers the collaboration inside a single model and takes smaller overhead. Extensive experiments on Cifar-10, Cifar-100, ImageNet32 and STL-10 corroborate the effectiveness of these four ways separately while combining them leads to further improvements. In particular, test errors on the STL-10 dataset are decreased by $9.28\%$ and $5.45\%$ for ResNet-18 and VGG-16 respectively. Moreover, our method is proven to be robust to label noise with experiments on Cifar-10 dataset. For example, our method has $3.53\%$ higher performance under $50\%$ noise ratio setting.

cs.CV↗

arXiv2020

TopologyGAN: Topology Optimization Using Generative Adversarial Networks Based on Physical Fields Over the Initial Domain

In topology optimization using deep learning, load and boundary conditions represented as vectors or sparse matrices often miss the opportunity to encode a rich view of the design problem, leading to less than ideal generalization results. We propose a new data-driven topology optimization model called TopologyGAN that takes advantage of various physical fields computed on the original, unoptimized material domain, as inputs to the generator of a conditional generative adversarial network (cGAN). Compared to a baseline cGAN, TopologyGAN achieves a nearly $3\times$ reduction in the mean squared error and a $2.5\times$ reduction in the mean absolute error on test problems involving previously unseen boundary conditions. Built on several existing network models, we also introduce a hybrid network called U-SE(Squeeze-and-Excitation)-ResNet for the generator that further increases the overall accuracy. We publicly share our full implementation and trained network.

cs.CE↗

arXiv2020

Broadband Ultrahigh-Resolution chip-scale Scanning Soliton Dual-Comb Spectroscopy

We present a chip-scale scanning dual-comb spectroscopy (SDCS) approach for broadband ultrahigh-resolution spectral acquisition. SDCS uses Si3N4 microring resonators that generate two single soliton micro-combs spanning 37 THz (300 nm) on the same chip from a single 1550-nm laser, forming a high-mutual-coherence dual-comb. We realize continuous tuning of the dual-comb system over the entire optical span of 37.5 THz with high precision using integrated microheater-based wavelength trackers. This continuous wavelength tuning is enabled by simultaneous tuning of the laser frequency and the two single soliton micro-combs over a full free spectral range of the microrings. We measure the SDCS resolution to be 319+-4.6 kHz. Using this SDCS system, we perform the molecular absorption spectroscopy of H13C14N over its 2.3 THz (18 nm)-wide overtone band, and show that the massively parallel heterodyning offered by the dual-comb expands the effective spectroscopic tuning speed of the laser by one order of magnitude. Our chip-scale SDCS opens the door to broadband spectrometry and massively parallel sensing with ultrahigh spectral resolution.

physics.app-ph↗

arXiv2019

Integrated near-field thermo-photovoltaics for on-demand heat recycling

The energy transferred via thermal radiation between two surfaces separated by nanometers distances (near-field) can be much larger than the blackbody limit. However, realizing a reconfigurable platform that utilizes this energy exchange mechanism to generate electricity in industrial and space applications on-demand, remains a challenge. The challenge lies in designing a platform that can separate two surfaces by a small and tunable gap while simultaneously maintaining a large temperature differential. Here, we present a fully integrated, reconfigurable and scalable platform operating in near-field regime that performs controlled heat extraction and energy recycling. Our platform relies on an integrated nano-electromechanical system (NEMS) that enables precise positioning of a large area thermal emitter within nanometers distances from a room-temperature germanium photodetector to form a thermo-photovoltaic (TPV) cell. We show over an order of magnitude higher power generation $\mathrm{P_{gen} \sim 1.25 \, μW \cdot cm^{-2}}$ from our TPV cell by tuning the gap between a hot emitter ($\mathrm{T_E \sim 880 \, K}$) and the cold photodetector ($\mathrm{T_D \sim 300 \, K}$) from $\mathrm{\sim 500 \, nm}$ to $\mathrm{\sim 100 \, nm}$. The significant enhancement in $\mathrm{P_{gen}}$ at such small distances is a clear indication of near-field heat transfer effect. Our electrostatically controlled NEMS switch consumes negligible tuning power ($\mathrm{P_{gen}/P_{NEMS} \sim 10^4}$) and relies on conventional silicon-based process technologies.

physics.app-ph↗

arXiv2012

Total Variation and Euler's Elastica for Supervised Learning

In recent years, total variation (TV) and Euler's elastica (EE) have been successfully applied to image processing tasks such as denoising and inpainting. This paper investigates how to extend TV and EE to the supervised learning settings on high dimensional data. The supervised learning problem can be formulated as an energy functional minimization under Tikhonov regularization scheme, where the energy is composed of a squared loss and a total variation smoothing (or Euler's elastica smoothing). Its solution via variational principles leads to an Euler-Lagrange PDE. However, the PDE is always high-dimensional and cannot be directly solved by common methods. Instead, radial basis functions are utilized to approximate the target function, reducing the problem to finding the linear coefficients of basis functions. We apply the proposed methods to supervised learning tasks (including binary classification, multi-class classification, and regression) on benchmark data sets. Extensive experiments have demonstrated promising results of the proposed methods.

cs.LG↗

arXiv2009

A feasibility study of treatment verification using EPID cine images for hypofractionated lung radiotherapy

We propose a novel approach for potential on-line treatment verification using cine EPID (Electronic Portal Imaging Device) images for hypofractionated lung radiotherapy based on a machine learning algorithm. Hypofractionated radiotherapy requires high precision. It is essential to effectively monitor the target to ensure that the tumor is within the beam aperture. We modeled the treatment verification problem as a two-class classification problem and applied an Artificial Neural Network (ANN) to classify the cine EPID images acquired during the treatment into corresponding classes with the tumor inside or outside of the beam aperture. Training samples were generated for the ANN using digitally reconstructed radiographs (DRRs) with artificially added shifts in tumor location to simulate cine EPID images with different tumor locations. Principal Component Analysis (PCA) was used to reduce the dimensionality of the training samples and cine EPID images acquired during the treatment. The proposed treatment verification algorithm was tested on five hypofractionated lung patients in a retrospective fashion. On average, our proposed algorithm achieved a 98.0% classification accuracy, a 97.6% recall rate, and a 99.7% precision rate.

physics.med-ph↗