Charts and atlases for nonlinear data-driven models of dynamics on manifolds

We introduce a method for learning minimal-dimensional dynamical models from high-dimensional time series data that lie on a low-dimensional manifold, as arises for many processes. For an arbitrary manifold, there is no smooth global coordinate representation, so following the formalism of differential topology we represent the manifold as an atlas of charts. We first partition the data into overlapping regions. Then undercomplete autoencoders are used to find low-dimensional coordinate representations for each region. We then use the data to learn dynamical models in each region, which together yield a global low-dimensional dynamical model. We apply this method to examples ranging from simple periodic dynamics to complex, nominally high-dimensional non-periodic bursting dynamics of the Kuramoto-Sivashinsky equation. We demonstrate that it: (1) can yield dynamical models of the lowest possible dimension, where previous methods generally cannot; (2) exhibits computational benefits including scalability, parallelizability, and adaptivity; and (3) separates state space into regions of distinct behaviours.

Adversarial Generation of Time-Frequency Features with application in audio synthesis

Time-frequency (TF) representations provide powerful and intuitive features for the analysis of time series such as audio. But still, generative modeling of audio in the TF domain is a subtle matter. Consequently, neural audio synthesis widely relies on directly modeling the waveform and previous attempts at unconditionally synthesizing audio from neurally generated TF features still struggle to produce audio at satisfying quality. In this contribution, focusing on the short-time Fourier transform, we discuss the challenges that arise in audio synthesis based on generated TF features and how to overcome them. We demonstrate the potential of deliberate generative TF modeling by training a generative adversarial network (GAN) on short-time Fourier features. We show that our TF-based network was able to outperform the state-of-the-art GAN generating waveform, despite the similar architecture in the two networks.

A Proactive Management Scheme for Data Synopses at the Edge

The combination of the infrastructure provided by the Internet of Things (IoT) with numerous processing nodes present at the Edge Computing (EC) ecosystem opens up new pathways to support intelligent applications. Such applications can be provided upon humongous volumes of data collected by IoT devices being transferred to the edge nodes through the network. Various processing activities can be performed on the discussed data and multiple collaborative opportunities between EC nodes can facilitate the execution of the desired tasks. In order to support an effective interaction between edge nodes, the knowledge about the geographically distributed data should be shared. Obviously, the migration of large amounts of data will harm the stability of the network stability and its performance. In this paper, we recommend the exchange of data synopses than real data between EC nodes to provide them with the necessary knowledge about peer nodes owning similar data. This knowledge can be valuable when considering decisions such as data/service migration and tasks offloading. We describe an continuous reasoning model that builds a temporal similarity map of the available datasets to get nodes understanding the evolution of data in their peers. We support the proposed decision making mechanism through an intelligent similarity extraction scheme based on an unsupervised machine learning model, and, at the same time, combine it with a statistical measure that represents the trend of the so-called discrepancy quantum. Our model can reveal the differences in the exchanged synopses and provide a datasets similarity map which becomes the appropriate knowledge base to support the desired processing activities. We present the problem under consideration and suggest a solution for that, while, at the same time, we reveal its advantages and disadvantages through a large number of experiments.

NetAdaptV2: Efficient Neural Architecture Search with Fast Super-Network Training and Architecture Optimization

Neural architecture search (NAS) typically consists of three main steps: training a super-network, training and evaluating sampled deep neural networks (DNNs), and training the discovered DNN. Most of the existing efforts speed up some steps at the cost of a significant slowdown of other steps or sacrificing the support of non-differentiable search metrics. The unbalanced reduction in the time spent per step limits the total search time reduction, and the inability to support non-differentiable search metrics limits the performance of discovered DNNs. In this paper, we present NetAdaptV2 with three innovations to better balance the time spent for each step while supporting non-differentiable search metrics. First, we propose channel-level bypass connections that merge network depth and layer width into a single search dimension to reduce the time for training and evaluating sampled DNNs. Second, ordered dropout is proposed to train multiple DNNs in a single forward-backward pass to decrease the time for training a super-network. Third, we propose the multi-layer coordinate descent optimizer that considers the interplay of multiple layers in each iteration of optimization to improve the performance of discovered DNNs while supporting non-differentiable search metrics. With these innovations, NetAdaptV2 reduces the total search time by up to $5.8\times$ on ImageNet and $2.4\times$ on NYU Depth V2, respectively, and discovers DNNs with better accuracy-latency/accuracy-MAC trade-offs than state-of-the-art NAS works. Moreover, the discovered DNN outperforms NAS-discovered MobileNetV3 by 1.8% higher top-1 accuracy with the same latency. The project website is http://netadapt.mit.edu.

A Deep Learning-based Quality Assessment and Segmentation System with a Large-scale Benchmark Dataset for Optical Coherence Tomographic Angiography Image

Optical Coherence Tomography Angiography (OCTA) is a non-invasive and non-contacting imaging technique providing visualization of microvasculature of retina and optic nerve head in human eyes in vivo. The adequate image quality of OCTA is the prerequisite for the subsequent quantification of retinal microvasculature. Traditionally, the image quality score based on signal strength is used for discriminating low quality. However, it is insufficient for identifying artefacts such as motion and off-centration, which rely specialized knowledge and need tedious and time-consuming manual identification. One of the most primary issues in OCTA analysis is to sort out the foveal avascular zone (FAZ) region in the retina, which highly correlates with any visual acuity disease. However, the variations in OCTA visual quality affect the performance of deep learning in any downstream marginally. Moreover, filtering the low-quality OCTA images out is both labor-intensive and time-consuming. To address these issues, we develop an automated computer-aided OCTA image processing system using deep neural networks as the classifier and segmentor to help ophthalmologists in clinical diagnosis and research. This system can be an assistive tool as it can process OCTA images of different formats to assess the quality and segment the FAZ area. The source code is freely available at https://github.com/shanzha09/COIPS.git. Another major contribution is the large-scale OCTA dataset, namely OCTA-25K-IQA-SEG we publicize for performance evaluation. It is comprised of four subsets, namely sOCTA-3$\times$3-10k, sOCTA-6$\times$6-14k, sOCTA-3$\times$3-1.1k-seg, and dOCTA-6$\times$6-1.1k-seg, which contains a total number of 25,665 images. The large-scale OCTA dataset is available at https://doi.org/10.5281/zenodo.5111975, https://doi.org/10.5281/zenodo.5111972.

Data-Driven Reduced-Order Modeling of Spatiotemporal Chaos with Neural Ordinary Differential Equations

Dissipative partial differential equations that exhibit chaotic dynamics tend to evolve to attractors that exist on finite-dimensional manifolds. We present a data-driven reduced order modeling method that capitalizes on this fact by finding the coordinates of this manifold and finding an ordinary differential equation (ODE) describing the dynamics in this coordinate system. The manifold coordinates are discovered using an undercomplete autoencoder -- a neural network (NN) that reduces then expands dimension. Then the ODE, in these coordinates, is approximated by a NN using the neural ODE framework. Both of these methods only require snapshots of data to learn a model, and the data can be widely and/or unevenly spaced. We apply this framework to the Kuramoto-Sivashinsky for different domain sizes that exhibit chaotic dynamics. With this system, we find that dimension reduction improves performance relative to predictions in the ambient space, where artifacts arise. Then, with the low-dimensional model, we vary the training data spacing and find excellent short- and long-time statistical recreation of the true dynamics for widely spaced data (spacing of ~0.7 Lyapunov times). We end by comparing performance with various degrees of dimension reduction, and find a "sweet spot" in terms of performance vs. dimension.

Follow the Curve: Robotic-Ultrasound Navigation with Learning Based Localization of Spinous Processes for Scoliosis Assessment

The scoliosis progression in adolescents requires close monitoring to timely take treatment measures. Ultrasound imaging is a radiation-free and low-cost alternative in scoliosis assessment to X-rays, which are typically used in clinical practice. However, ultrasound images are prone to speckle noises, making it challenging for sonographers to detect bony features and follow the spine's curvature. This paper introduces a robotic-ultrasound approach for spinal curvature tracking and automatic navigation. A fully connected network with deconvolutional heads is developed to locate the spinous process efficiently with real-time ultrasound images. We use this machine learning-based method to guide the motion of the robot-held ultrasound probe and follow the spinal curvature while capturing ultrasound images and correspondent position. We developed a new force-driven controller that automatically adjusts the probe's pose relative to the skin surface to ensure a good acoustic coupling between the probe and skin. After the scanning, the acquired data is used to reconstruct the coronal spinal image, where the deformity of the scoliosis spine can be assessed and measured. To evaluate the performance of our methodology, we conducted an experimental study with human subjects where the deviations from the image center during the robotized procedure are compared to that obtained from manual scanning. The angles of spinal deformity measured on spinal reconstruction images were similar for both methods, implying that they equally reflect human anatomy.

Scalable Spatiotemporally Varying Coefficient Modeling with Bayesian Kernelized Tensor Regression

As a regression technique in spatial statistics, spatiotemporally varying coefficient model (STVC) is an important tool to discover nonstationary and interpretable response-covariate associations over both space and time. However, it is difficult to apply STVC for large-scale spatiotemporal analysis due to the high computational cost. To address this challenge, we summarize the spatiotemporally varying coefficients using a third-order tensor structure and propose to reformulate the spatiotemporally varying coefficient model as a special low-rank tensor regression problem. The low-rank decomposition can effectively model the global patterns of the large data with substantially reduced number of parameters. To further incorporate the local spatiotemporal dependencies among the samples, we place Gaussian process (GP) priors on the spatial and temporal factor matrices to better encode local spatial and temporal processes on each factor component. We refer to the overall framework as Bayesian Kernelized Tensor Regression (BKTR). For model inference, we develop an efficient Markov chain Monte Carlo (MCMC) algorithm, which uses Gibbs sampling to update factor matrices and slice sampling to update kernel hyperparameters. We conduct extensive experiments on both synthetic and real-world data sets, and our results confirm the superior performance and efficiency of BKTR for model estimation and parameter inference.

Sound Event Detection in Multichannel Audio using Convolutional Time-Frequency-Channel Squeeze and Excitation

In this study, we introduce a convolutional time-frequency-channel "Squeeze and Excitation" (tfc-SE) module to explicitly model inter-dependencies between the time-frequency domain and multiple channels. The tfc-SE module consists of two parts: tf-SE block and c-SE block which are designed to provide attention on time-frequency and channel domain, respectively, for adaptively recalibrating the input feature map. The proposed tfc-SE module, together with a popular Convolutional Recurrent Neural Network (CRNN) model, are evaluated on a multi-channel sound event detection task with overlapping audio sources: the training and test data are synthesized TUT Sound Events 2018 datasets, recorded with microphone arrays. We show that the tfc-SE module can be incorporated into the CRNN model at a small additional computational cost and bring significant improvements on sound event detection accuracy. We also perform detailed ablation studies by analyzing various factors that may influence the performance of the SE blocks. We show that with the best tfc-SE block, error rate (ER) decreases from 0.2538 to 0.2026, relative 20.17\% reduction of ER, and 5.72\% improvement of F1 score. The results indicate that the learned acoustic embeddings with the tfc-SE module efficiently strengthen time-frequency and channel-wise feature representations to improve the discriminative performance.

Lifelong Infinite Mixture Model Based on Knowledge-Driven Dirichlet Process

Recent research efforts in lifelong learning propose to grow a mixture of models to adapt to an increasing number of tasks. The proposed methodology shows promising results in overcoming catastrophic forgetting. However, the theory behind these successful models is still not well understood. In this paper, we perform the theoretical analysis for lifelong learning models by deriving the risk bounds based on the discrepancy distance between the probabilistic representation of data generated by the model and that corresponding to the target dataset. Inspired by the theoretical analysis, we introduce a new lifelong learning approach, namely the Lifelong Infinite Mixture (LIMix) model, which can automatically expand its network architectures or choose an appropriate component to adapt its parameters for learning a new task, while preserving its previously learnt information. We propose to incorporate the knowledge by means of Dirichlet processes by using a gating mechanism which computes the dependence between the knowledge learnt previously and stored in each component, and a new set of data. Besides, we train a compact Student model which can accumulate cross-domain representations over time and make quick inferences. The code is available at https://github.com/dtuzi123/Lifelong-infinite-mixture-model.