Neural Architecture Search (NAS) has seen an explosion of research in the past few years. A variety of methods have been proposed to perform NAS, including reinforcement learning, Bayesian optimization with a Gaussian process model, evolutionary search, and gradient descent. In this work, we design a NAS algorithm that performs Bayesian optimization using a neural network model. We develop a path-based encoding scheme to featurize the neural architectures that are used to train the neural network model. This strategy is particularly effective for encoding architectures in cell-based search spaces. After training on just 200 random neural architectures, we are able to predict the validation accuracy of a new architecture to within one percent of its true accuracy on average, for popular search spaces. This may be of independent interest beyond Bayesian neural architecture search. We test our algorithm on the NASBench (Ying et al. 2019) and DARTS (Liu et al. 2018) search spaces, and we show that our algorithm outperforms other NAS methods including evolutionary search, reinforcement learning, AlphaX, ASHA, and DARTS. Our algorithm is over 100x more efficient than random search, and 3.8x more efficient than the next-best algorithm on the NASBench dataset. As there have been problems with fair and reproducible experimental evauations in the field of NAS, we adhere to the recent NAS research checklist (Lindauer and Hutter 2019) to facilitate NAS research. In particular, our implementation has been made publicly available, including all details needed to fully reproduce our results.

Car following models have been widely applied and made remarkable achievements in traffic engineering. However, the traffic micro-simulation accuracy of car following models in a platoon level, especially during traffic oscillations, still needs to be enhanced. Rather than using traditional individual car following models, we proposed a new trajectory generation approach to generate platoon level trajectories given the first leading vehicle's trajectory. In this paper, we discussed the temporal and spatial error propagation issue for the traditional approach by a car following block diagram representation. Based on the analysis, we pointed out that error comes from the training method and the model structure. In order to fix that, we adopt two improvements on the basis of the traditional LSTM based car following model. We utilized a scheduled sampling technique during the training process to solve the error propagation in the temporal dimension. Furthermore, we developed a unidirectional interconnected LSTM model structure to extract trajectories features from the perspective of the platoon. As indicated by the systematic empirical experiments, the proposed novel structure could efficiently reduce the temporal and spatial error propagation. Compared with the traditional LSTM based car following model, the proposed model has almost 40% less error. The findings will benefit the design and analysis of micro-simulation for platoon level car following models.

Differentiable neural architecture search has been a popular methodology of exploring architectures for deep learning. Despite the great advantage of search efficiency, it often suffers weak stability, which obstacles it from being applied to a large search space or being flexibly adjusted to different scenarios. This paper investigates DARTS, the currently most popular differentiable search algorithm, and points out an important factor of instability, which lies in its approximation on the gradients of architectural parameters. Based on this analysis, we propose an amending term for computing architectural gradients by making use of a direct property of the opti-mality of network parameter optimization. Our approach mathematically guarantees that gradient estimation follows a roughly correct direction, which leads the search stage to converge on reasonable architectures. In practice, our algorithm is easily implemented and added to DARTSbased approaches efficiently. Experiments on CIFAR and ImageNet demonstrate that our approach enjoys accuracy gain and, more importantly, enables DARTSbased approaches to explore much larger search spaces that have not been studied before. Neural architecture search (NAS) has been an important topic in the research area of automated machine learning (AutoML). The idea is to replace the manual way of designing neural network architectures with an automatic algorithm, by which deep learning methods become more flexible in fitting complex data distributions, e.g., large-scale image datasets.

There is no consensus yet on the question whether adaptive gradient methods like Adam are easier to use than non-adaptive optimization methods like SGD. In this work, we fill in the important, yet ambiguous concept of `ease-of-use' by defining an optimizer's \emph{tunability}: How easy is it to find good hyperparameter configurations using automatic random hyperparameter search? We propose a practical and universal quantitative measure for optimizer tunability that can form the basis for a fair optimizer benchmark. Evaluating a variety of optimizers on an extensive set of standard datasets and architectures, we find that Adam is the most tunable for the majority of problems, especially with a low budget for hyperparameter tuning.

In this paper, we propose the idea of radial scaling in frequency domain and activation functions with compact support to produce a multi-scale DNN (MscaleDNN), which will have the multi-scale capability in approximating high frequency and high dimensional functions and speeding up the solution of high dimensional PDEs. Numerical results on high dimensional function fitting and solutions of high dimensional PDEs, using loss functions with either Ritz energy or least squared PDE residuals, have validated the increased power of multi-scale resolution and high frequency capturing of the proposed MscaleDNN.

End-to-end performance estimation and measurement of deep neural network (DNN) systems become more important with increasing complexity of DNN systems consisting of hardware and software components. The methodology proposed in this paper aims at a reduced turn-around time for evaluating different design choices of hardware and software components of DNN systems. This reduction is achieved by moving the performance estimation from the implementation phase to the concept phase by employing virtual hardware models instead of gathering measurement results from physical prototypes. Deep learning compilers introduce hardware-specific transformations and are, therefore, considered a part of the design flow of virtual system models to extract end-to-end performance estimations. To validate the run-time accuracy of the proposed methodology, a system processing the DilatedVGG DNN is realized both as virtual system model and as hardware implementation. The results show that up to 92 % accuracy can be reached in predicting the processing time of the DNN inference.

Applications in quantitative finance such as optimal trade execution, risk management of options, and optimal asset allocation involve the solution of high dimensional and nonlinear Partial Differential Equations (PDEs). The connection between PDEs and systems of Forward-Backward Stochastic Differential Equations (FBSDEs) enables the use of advanced simulation techniques to be applied even in the high dimensional setting. Unfortunately, when the underlying application contains nonlinear terms, then classical methods both for simulation and numerical methods for PDEs suffer from the curse of dimensionality. Inspired by the success of deep learning, several researchers have recently proposed to address the solution of FBSDEs using deep learning. We discuss the dynamical systems point of view of deep learning and compare several architectures in terms of stability, generalization, and robustness. In order to speed up the computations, we propose to use a multilevel discretization technique. Our preliminary results suggest that the multilevel discretization method improves solutions times by an order of magnitude compared to existing methods without sacrificing stability or robustness.

Training neural networks on image datasets generally require extensive experimentation to find the optimal learning rate regime. Especially, for the cases of adversarial training or for training a newly synthesized model, one would not know the best learning rate regime beforehand. We propose an automated algorithm for determining the learning rate trajectory, that works across datasets and models for both natural and adversarial training, without requiring any dataset/model specific tuning. It is a stand-alone, parameterless, adaptive approach with no computational overhead. We theoretically discuss the algorithm's convergence behavior. We empirically validate our algorithm extensively. Our results show that our proposed approach \emph{consistently} achieves top-level accuracy compared to SOTA baselines in the literature in natural as well as adversarial training.

Autoregressive sequence models achieve state-of-the-art performance in domains like machine translation. However, due to the autoregressive factorization nature, these models suffer from heavy latency during inference. Recently, non-autoregressive sequence models were proposed to speed up the inference time. However, these models assume that the decoding process of each token is conditionally independent of others. Such a generation process sometimes makes the output sentence inconsistent, and thus the learned non-autoregressive models could only achieve inferior accuracy compared to their autoregressive counterparts. To improve then decoding consistency and reduce the inference cost at the same time, we propose to incorporate a structured inference module into the non-autoregressive models. Specifically, we design an efficient approximation for Conditional Random Fields (CRF) for non-autoregressive sequence models, and further propose a dynamic transition technique to model positional contexts in the CRF. Experiments in machine translation show that while increasing little latency (8~14ms), our model could achieve significantly better translation performance than previous non-autoregressive models on different translation datasets. In particular, for the WMT14 En-De dataset, our model obtains a BLEU score of 26.80, which largely outperforms the previous non-autoregressive baselines and is only 0.61 lower in BLEU than purely autoregressive models.

Predicting the relationship between a molecule's structure and its odor remains a difficult, decades-old task. This problem, termed quantitative structure-odor relationship (QSOR) modeling, is an important challenge in chemistry, impacting human nutrition, manufacture of synthetic fragrance, the environment, and sensory neuroscience. We propose the use of graph neural networks for QSOR, and show they significantly out-perform prior methods on a novel data set labeled by olfactory experts. Additional analysis shows that the learned embeddings from graph neural networks capture a meaningful odor space representation of the underlying relationship between structure and odor, as demonstrated by strong performance on two challenging transfer learning tasks. Machine learning has already had a large impact on the senses of sight and sound. Based on these early results with graph neural networks for molecular properties, we hope machine learning can eventually do for olfaction what it has already done for vision and hearing.