There are many sensor fusion frameworks proposed in the literature using different sensors and fusion methods combinations and configurations. More focus has been on improving the accuracy performance; however, the implementation feasibility of these frameworks in an autonomous vehicle is less explored. Some fusion architectures can perform very well in lab conditions using powerful computational resources; however, in real-world applications, they cannot be implemented in an embedded edge computer due to their high cost and computational need. We propose a new hybrid multi-sensor fusion pipeline configuration that performs environment perception for autonomous vehicles such as road segmentation, obstacle detection, and tracking. This fusion framework uses a proposed encoder-decoder based Fully Convolutional Neural Network (FCNx) and a traditional Extended Kalman Filter (EKF) nonlinear state estimator method. It also uses a configuration of camera, LiDAR, and radar sensors that are best suited for each fusion method. The goal of this hybrid framework is to provide a cost-effective, lightweight, modular, and robust (in case of a sensor failure) fusion system solution. It uses FCNx algorithm that improve road detection accuracy compared to benchmark models while maintaining real-time efficiency that can be used in an autonomous vehicle embedded computer. Tested on over 3K road scenes, our fusion algorithm shows better performance in various environment scenarios compared to baseline benchmark networks. Moreover, the algorithm is implemented in a vehicle and tested using actual sensor data collected from a vehicle, performing real-time environment perception.

The recent trend in adapting ultra-energy-efficient (but error-prone) nanomagnetic devices to non-Boolean computing and information processing (e.g. stochastic/probabilistic computing, neuromorphic, belief networks, etc) has resulted in rapid strides in new computing modalities. Of particular interest are Bayesian networks (BN) which may see revolutionary advances when adapted to a specific type of nanomagnetic devices. Here, we develop a novel nanomagnet-based computing substrate for BN that allows high-speed sampling from an arbitrary Bayesian graph. We show that magneto-tunneling junctions (MTJs) can be used for electrically programmable ‘sub-nanosecond’ probability sample generation by co-optimizing voltage-controlled magnetic anisotropy and spin transfer torque. We also discuss that just by engineering local magnetostriction in the soft layers of MTJs, one can stochastically couple them for programmable conditional sample generation as well. This obviates the need for extensive energy-inefficient hardware like OP-AMPS, gates, shift-registers, etc to generate the correlations. Based on the above findings, we present an architectural design and computation flow of the MTJ network to map an arbitrary Bayesian graph where we develop circuits to program and induce switching and interactions among MTJs. Our discussed framework can lead to a new generation of stochastic computing hardware for various other computing models, such as stochastic programming and Bayesian deep learning. This can spawn a novel genre of ultra-energy-efficient, extremely powerful computing paradigms, which is a transformational advance.

The treatment of mood disorders, which can become a lifelong process, varies widely in efficacy between individuals. Most options to monitor mood rely on subjective self-reports and clinical visits, which can be burdensome and may not portray an accurate representation of what the individual is experiencing. A passive method to monitor mood could be a useful tool for those with these disorders. Some previously proposed models utilized sensors from smartphones and wearables, such as the accelerometer. This study examined a novel approach of processing accelerometer data collected from smartphones only while participants of the open-science branch of the BiAffect study were typing. The data were modeled by von Mises-Fisher distributions and weighted networks to identify clusters relating to different typing positions unique for each participant. Longitudinal features were derived from the clustered data and used in machine learning models to predict clinically relevant changes in depression from clinical and typing measures. Model accuracy was approximately 95%, with 97% area under the ROC curve (AUC). The accelerometer features outperformed the vast majority of clinical and typing features, which suggested that this new approach to analyzing accelerometer data could contribute towards unobtrusive detection of changes in depression severity without the need for clinical input.

This work proposes a novel Energy-aware Network Operator Search (ENOS) approach to address the energy-accuracy trade-offs of a deep neural network (DNN) accelerator. In recent years, novel hardware-friendly inference operators such as binary-weight, multiplication-free, and deep-shift have been proposed to improve the computational efficiency of a DNN accelerator. However, simplifying DNN operators invariably comes at lower accuracy, especially on complex processing tasks. While prior works generally implement the same inference operator throughout the neural architecture, the proposed ENOS framework allows an optimal layer-wise integration of inference operators with optimal precision to maintain high prediction accuracy and high energy efficiency. The search in ENOS is formulated as a continuous optimization problem, solvable using gradient descent methods, thereby minimally increasing the training cost when learning both layer-wise inference operators and weights. Utilizing ENOS, we discuss multiply-accumulate (MAC) cores for digital spatial architectures that can be reconfigured to different operators and varying computing precision. ENOS training methods with single and bi-level optimization objectives are discussed and compared. We also discuss a sequential operator assignment strategy in ENOS that only learns the assignment for one layer in one training step. Furthermore, a stochastic mode of ENOS is also presented. ENOS is characterized on ShuffleNet and SqueezeNet using CIFAR10 and CIFAR100. Compared to the conventional uni-operator approaches, under the same energy budget, ENOS improves accuracy by 10–20%. ENOS also outperforms the accuracy of comparable mixed-precision uni-operator implementations by 3-5% for the same energy budget.

Background: Passive sensing through smartphone keyboard data can be used to identify and monitor symptoms of mood disorders with low participant burden. Behavioral phenotyping based on mobile keystroke data can aid in clinical decision-making and provide insights into the individual symptoms of mood disorders. Objective: This study aims to derive digital phenotypes based on smartphone keyboard backspace use among 128 community adults across 2948 observations using a Bayesian mixture model. Methods: Eligible study participants completed a virtual screening visit where all eligible participants were instructed to download the custom-built BiAffect smartphone keyboard (University of Illinois). The BiAffect keyboard unobtrusively captures keystroke dynamics. All eligible and consenting participants were instructed to use this keyboard exclusively for up to 4 weeks of the study in real life, and participants’ compliance was checked at the 2 follow-up visits at week 2 and week 4. As part of the research protocol, every study participant underwent evaluations by a study psychiatrist during each visit. Results: We found that derived phenotypes were associated with not only the diagnoses and severity of depression and mania but also specific individual symptoms. Using a linear mixed-effects model with random intercepts accounting for the nested data structure from daily data, the backspace rates on the continuous scale did not differ between participants in the healthy control and in the mood disorders groups (P=.11). The 3-class model had mean backspace rates of 0.112, 0.180, and 0.268, respectively, with a SD of 0.048. In total, 3 classes, respectively, were estimated to comprise 37.5% (n=47), 54.4% (n=72), and 8.1% (n=9) of the sample. We grouped individuals into Low, Medium, and High backspace rate groups. Individuals with unipolar mood disorder were predominantly in the Medium group (n=54), with some in the Low group (n=27) and a few in the High group (n=6). The Medium group, compared with the Low group, had significantly higher ratings of depression (b=2.32, P=.008). The High group was not associated with ratings of depression with (P=.88) or without (P=.27) adjustment for medication and diagnoses. The High group, compared with the Low group, was associated with both nonzero ratings (b=1.91, P=.02) and higher ratings of mania (b=1.46, P<.001). The High group, compared with the Low group, showed significantly higher odds of elevated mood (P=.03), motor activity (P=.04), and irritability (P<.05). Conclusions: This study demonstrates the promise of mobile typing kinematics in mood disorder research and practice. Monitoring a single mobile typing kinematic feature, that is, backspace rates, through passive sensing imposes a low burden on the participants. Based on real-life keystroke data, our derived digital phenotypes from this single feature can be useful for researchers and practitioners to distinguish between individuals with and those without mood disorder symptoms.