AutoTransfer: Subject Transfer Learning with Censored Representations on Biosignals Data.

We investigate a regularization framework for subject transfer learning in which we train an encoder and classifier to minimize classification loss, subject to a penalty measuring independence between the latent representation and the subject label. We introduce three notions of independence and corresponding penalty terms using mutual information or divergence as a proxy for independence. For each penalty term, we provide several concrete estimation algorithms, using analytic methods as well as neural critic functions. We propose a hands-off strategy for applying this diverse family of regularization schemes to a new dataset, which we call "Auto Transfer". We evaluate the performance of these individual regularization strategies under our AutoTransfer framework on EEG, EMG, and ECoG datasets, showing that these approaches can improve subject transfer learning for challenging real-world datasets.

EEG-GAT: Graph Attention Networks for Classification of Electroencephalogram (EEG) Signals.

Graph neural networks (GNN) are an emerging framework in the deep learning community. In most GNN applications, the graph topology of data samples is provided in the dataset. Specifically, the graph shift operator (GSO), which could be adjacency, graph Laplacian, or their normalizations, is known a priori. However we often have no knowledge of the grand-truth graph topology underlying real-world datasets. One example of this is to extract subject-invariant features from physiological electroencephalogram (EEG) to predict a cognitive task. Previous methods use electrode sites to represent a node in the graph and connect them in various ways to hand-engineer a GSO e.g., i) each pair of electrode sites is connected to form a complete graph, ii) a specific number of electrode sites are connected to form a k-nearest neighbor graph, iii) each pair of electrode site is connected only if the Euclidean distance is within a heuristic threshold. In this paper, we overcome this limitation by parameterizing the GSO using a multi-head attention mechanism to explore the functional neural connectivity subject to a cognitive task between different electrode sites, and simultaneously learn the unsupervised graph topology in conjunction with the parameters of graph convolutional kernels.

Inverse regular perturbation with ML-assisted phasor correction for fiber nonlinearity compensation.

We improve an inverse regular perturbation (RP) model using a machine learning (ML) technique. The proposed learned RP (LRP) model jointly optimizes step-size, gain and phase rotation for individual RP branches. We demonstrate that the proposed LRP can outperform the corresponding learned digital back-propagation (DBP) method based on a split-step Fourier method (SSFM), with up to 0.75 dB gain in a 800 km standard single mode fiber link. Our LRP also allows a fractional step-per-span (SPS) modeling to reduce complexity while maintaining superior performance over a 1-SPS SSFM-DBP.

DNN-assisted phase distance tuned PSK modulation for PAM4-to-QPSK format conversion gateway node.

This paper proposes an optical gateway that converts pulse amplitude modulation (PAM) format to phase shift keying (PSK) modulation format, enabling flexible intensity-to-phase mapping without relying on optical-electrical-optical data conversion at heterogenous network connections. A proof-of-principle experiment shows that optically converted PSK signals from regular PAM signals will induce non-uniform irregular phase noise distortion. The proposed optical gateway is designed to provide an optimized phase distance for PSK signals such that an achievable information rate is maximized by a deep learning-based decision on the receiver side. The phase distance-tuned PSK signals offer performance improvement of about 4 dB gain at a target generalized mutual information by making use of a digital neural network recovery.

EEG-GNN: Graph Neural Networks for Classification of Electroencephalogram (EEG) Signals.

Convolutional neural networks (CNN) have been frequently used to extract subject-invariant features from electroencephalogram (EEG) for classification tasks. This approach holds the underlying assumption that electrodes are equidistant analogous to pixels of an image and hence fails to explore/exploit the complex functional neural connectivity between different electrode sites. We overcome this limitation by tailoring the concepts of convolution and pooling applied to 2D grid-like inputs for the functional network of electrode sites. Furthermore, we develop various graph neural network (GNN) models that project electrodes onto the nodes of a graph, where the node features are represented as EEG channel samples collected over a trial, and nodes can be connected by weighted/unweighted edges according to a flexible policy formulated by a neuroscientist. The empirical evaluations show that our proposed GNN-based framework outperforms standard CNN classifiers across ErrP, and RSVP datasets, as well as allowing neuroscientific interpretability and explainability to deep learning methods tailored to EEG related classification problems. Another practical advantage of our GNN-based framework is that it can be used in EEG channel selection, which is critical for reducing computational cost, and designing portable EEG headsets.

Universal Physiological Representation Learning With Soft-Disentangled Rateless Autoencoders.

Human computer interaction (HCI) involves a multidisciplinary fusion of technologies, through which the control of external devices could be achieved by monitoring physiological status of users. However, physiological biosignals often vary across users and recording sessions due to unstable physical/mental conditions and task-irrelevant activities. To deal with this challenge, we propose a method of adversarial feature encoding with the concept of a Rateless Autoencoder (RAE), in order to exploit disentangled, nuisance-robust, and universal representations. We achieve a good trade-off between user-specific and task-relevant features by making use of the stochastic disentanglement of the latent representations by adopting additional adversarial networks. The proposed model is applicable to a wider range of unknown users and tasks as well as different classifiers. Results on cross-subject transfer evaluations show the advantages of the proposed framework, with up to an 11.6% improvement in the average subject-transfer classification accuracy.

Spectrally sparse optical coherence tomography.

Swept-source optical coherence tomography (OCT) typically relies on expensive and complex swept-source lasers, the cost of which currently limits the suitability of OCT for new applications. In this work, we demonstrate spectrally sparse OCT utilizing randomly spaced low-bandwidth optical chirps, suitable for low-cost implementation with telecommunications grade devices. Micron scale distance estimation accuracy with a resolution of 40 µm at a standoff imaging distance greater than 10 cm is demonstrated using a stepped chirp approach with approximately 23% occupancy of 4 THz bandwidth. For imaging of sparse scenes, comparable performance to full bandwidth occupancy is verified for metallic targets.

Disentangled Adversarial Transfer Learning for Physiological Biosignals.

Recent developments in wearable sensors demonstrate promising results for monitoring physiological status in effective and comfortable ways. One major challenge of physiological status assessment is the problem of transfer learning caused by the domain inconsistency of biosignals across users or different recording sessions from the same user. We propose an adversarial inference approach for transfer learning to extract disentangled nuisance-robust representations from physiological biosignal data in stress status level assessment. We exploit the trade-off between task-related features and person-discriminative information by using both an adversary network and a nuisance network to jointly manipulate and disentangle the learned latent representations by the encoder, which are then input to a discriminative classifier. Results on cross-subjects transfer evaluations demonstrate the benefits of the proposed adversarial framework, and thus show its capabilities to adapt to a broader range of subjects. Finally we highlight that our proposed adversarial transfer learning approach is also applicable to other deep feature learning frameworks.

Proposal and Evaluation of Visual Haptics for Manipulation of Remote Machine System.

Remote machine systems have drawn a lot of attention owing to accelerations of virtual reality (VR), augmented reality (AR), and the fifth generation (5G) networks. Despite recent trends of developing autonomous systems, the realization of sophisticated dexterous hand that can fully replace human hands is considered to be decades away. It is also extremely difficult to reproduce the sensilla of complex human hands. On the other hand, it is known that humans can perceive haptic information from visual information even without any physical feedback as cross modal sensation between visual and haptics sensations or pseudo haptics. In this paper, we propose a visual haptic technology, where haptic information is visualized in more perceptual images overlaid at the contact points of a remote machine hand. The usability of the proposed visual haptics was evaluated by subject's brain waves aiming to find out a new approach for quantifying "sense of oneness." In our proof-of-concept experiments using VR, subjects are asked to operate a virtual arm and hand presented in the VR space, and the performance of the operation with and without visual haptics information as measured with brain wave sensing. Consequently, three results were verified. Firstly, the information flow in the brain were significantly reduced with the proposed visual haptics for the whole α, ß, and θ-waves by 45% across nine subjects. This result suggests that superimposing visual effects may be able to reduce the cognitive burden on the operator during the manipulation for the remote machine system. Secondly, high correlation (Pearson correlation factor of 0.795 at a p-value of 0.011) was verified between the subjective usability points and the brainwave measurement results. Finally, the number of the task successes across sessions were improved in the presence of overlaid visual stimulus. It implies that the visual haptics image could also facilitate operators' pre-training to get skillful at manipulating the remote machine interface more quickly.

Disentangled Adversarial Autoencoder for Subject-Invariant Physiological Feature Extraction.

Recent developments in biosignal processing have enabled users to exploit their physiological status for manipulating devices in a reliable and safe manner. One major challenge of physiological sensing lies in the variability of biosignals across different users and tasks. To address this issue, we propose an adversarial feature extractor for transfer learning to exploit disentangled universal representations. We consider the trade-off between task-relevant features and user-discriminative information by introducing additional adversary and nuisance networks in order to manipulate the latent representations such that the learned feature extractor is applicable to unknown users and various tasks. Results on cross-subject transfer evaluations exhibit the benefits of the proposed framework, with up to 8.8% improvement in average accuracy of classification, and demonstrate adaptability to a broader range of subjects.

Learning Invariant Representations from EEG via Adversarial Inference.

Discovering and exploiting shared, invariant neural activity in electroencephalogram (EEG) based classification tasks is of significant interest for generalizability of decoding models across subjects or EEG recording sessions. While deep neural networks are recently emerging as generic EEG feature extractors, this transfer learning aspect usually relies on the prior assumption that deep networks naturally behave as subject- (or session-) invariant EEG feature extractors. We propose a further step towards invariance of EEG deep learning frameworks in a systemic way during model training. We introduce an adversarial inference approach to learn representations that are invariant to inter-subject variabilities within a discriminative setting. We perform experimental studies using a publicly available motor imagery EEG dataset, and state-of-the-art convolutional neural network based EEG decoding models within the proposed adversarial learning framework. We present our results in cross-subject model transfer scenarios, demonstrate neurophysiological interpretations of the learned networks, and discuss potential insights offered by adversarial inference to the growing field of deep learning for EEG.

Adversarial Deep Learning in EEG Biometrics.

Deep learning methods for person identification based on electroencephalographic (EEG) brain activity encounters the problem of exploiting the temporally correlated structures or recording session specific variability within EEG. Furthermore, recent methods have mostly trained and evaluated based on single session EEG data. We address this problem from an invariant representation learning perspective. We propose an adversarial inference approach to extend such deep learning models to learn session-invariant person-discriminative representations that can provide robustness in terms of longitudinal usability. Using adversarial learning within a deep convolutional network, we empirically assess and show improvements with our approach based on longitudinally collected EEG data for person identification from half-second EEG epochs.

Deep Neural Network Inverse Design of Integrated Photonic Power Splitters.

Predicting physical response of an artificially structured material is of particular interest for scientific and engineering applications. Here we use deep learning to predict optical response of artificially engineered nanophotonic devices. In addition to predicting forward approximation of transmission response for any given topology, this approach allows us to inversely approximate designs for a targeted optical response. Our Deep Neural Network (DNN) could design compact (2.6 × 2.6 µm2) silicon-on-insulator (SOI)-based 1 × 2 power splitters with various target splitting ratios in a fraction of a second. This model is trained to minimize the reflection (to smaller than ~ -20 dB) while achieving maximum transmission efficiency above 90% and target splitting specifications. This approach paves the way for rapid design of integrated photonic components relying on complex nanostructures.

High-accuracy user identification using EEG biometrics.

We analyze brain waves acquired through a consumer-grade EEG device to investigate its capabilities for user identification and authentication. First, we show the statistical significance of the P300 component in event-related potential (ERP) data from 14-channel EEGs across 25 subjects. We then apply a variety of machine learning techniques, comparing the user identification performance of various different combinations of a dimensionality reduction technique followed by a classification algorithm. Experimental results show that an identification accuracy of 72% can be achieved using only a single 800 ms ERP epoch. In addition, we demonstrate that the user identification accuracy can be significantly improved to more than 96.7% by joint classification of multiple epochs.

An MMI-based wavelength combiner employing non-uniform refractive index distribution.

A novel wavelength combiner using non-uniform refractive index distribution within a multimode interference device is proposed and simulated. The refractive index step creates separate localized modes with different effective refractive indices and two modes are strongly excited which form the basis of an interferometer. We applied the concept to 1.30/1.31 µm and 1.31/1.55 µm wavelength combiners on an InP substrate. The lengths of the devices are 1272 µm and 484 µm with simulated insertion losses of 0.6 dB and 0.67 dB respectively.

High-dimensional modulation for coherent optical communications systems.

In this paper, we examine the performance of several modulation formats in more than four dimensions for coherent optical communications systems. We compare two high-dimensional modulation design methodologies based on spherical cutting of lattices and block coding of a 'base constellation' of binary phase shift keying (BPSK) on each dimension. The performances of modulation formats generated with these methodologies is analyzed in the asymptotic signal-to-noise ratio regime and for an additive white Gaussian noise (AWGN) channel. We then study the application of both types of high-dimensional modulation formats to standard single-mode fiber (SSMF) transmission systems. For modulation with spectral efficiencies comparable to dual-polarization (DP-) BPSK, polarization-switched quaternary phase shift keying (PS-QPSK) and DP-QPSK, we demonstrate SNR gains of up to 3 dB, 0.9 dB and 1 dB respectively, at a BER of 10(-3).

An MMI-based polarization splitter using patterned metal and tilted joint.

A novel polarization splitter on an InP substrate utilizing an MMI coupler loaded with a dielectric and gold layer pad is proposed and simulated. A tilted joint is used for adjusting the phases of TE and TM modes. The MMI section is less than 540 µm. Simulations show that the device has a polarization extinction ratio over 23 dB and an insertion loss below 0.7 dB over the entire C-band for both TE and TM polarizations. The device design was optimized to maximize the wavelength range and tolerance for manufacturing variations.

High-order statistical equalizer for nonlinearity compensation in dispersion-managed coherent optical communications.

Fiber nonlinearity has become a major limiting factor to realize ultra-high-speed optical communications. We propose a fractionally-spaced equalizer which exploits a trained high-order statistics to deal with data-pattern dependent nonlinear impairments in fiber-optic communications. The computer simulation reveals that the proposed 3-tap equalizer improves Q-factor by more than 2 dB for long-haul transmissions of 5,230 km distance and 40 Gbps data rate. We also demonstrate that the joint use of a digital backpropagation (DBP) and the proposed equalizer offers an additional 1-2 dB performance improvement due to the channel shortening gain. A performance in high-speed transmissions of 100 Gbps and beyond is evaluated as well.

Mode-evolution-based polarization rotator-splitter design via simple fabrication process.

A mode-evolution-based polarization rotator-splitter built on InP substrate is proposed by combining a mode converter and an adiabatic asymmetric Y-coupler. The mode converter, consisting of a bi-level taper and a width taper, effectively converts the fundamental TM mode into the second order TE mode without changing the polarization of the fundamental TE mode. The following adiabatic asymmetric Y-coupler splits the fundamental and the second order TE modes and also converts the second order TE mode into the fundamental TE mode. A shallow etched structure is proposed for the width taper to enhance the polarization conversion efficiency. The device has a total length of 1350 µm, a polarization extinction ratio over 25 dB and an insertion loss below 0.5 dB both for TE and TM modes, over the wavelength range from 1528 to 1612 nm covering all C + L band. Because the device is designed based on mode evolution principle, it has a large fabrication tolerance. The insertion loss remains below 1 dB and the polarization extinction ratio remains over 17 dB with respect to a width variation of +/- 0.12 µm at the wavelength of 1570 nm, or +/- 0.08 µm over the entire C + L band.