We consider multi-level composite optimization problems where each mapping in the composition is the expectation over a family of random smooth mappings or the sum of some finite number of smooth mappings. We present a normalized proximal approximate gradient (NPAG) method where the approximate gradients are obtained via nested stochastic variance reduction. In order to find an approximate stationary point where the expected norm of its gradient mapping is less than $\epsilon$, the total sample complexity of our method is $O(\epsilon^{-3})$ in the expectation case, and $O(N+\sqrt{N}\epsilon^{-2})$ in the finite-sum case where $N$ is the total number of functions across all composition levels. In addition, the dependence of our total sample complexity on the number of composition levels is polynomial, rather than exponential as in previous work.

We describe a method for selecting relevant new training data for the LSTM-based domain selection component of our personal assistant system. Adding more annotated training data for any ML system typically improves accuracy, but only if it provides examples not already adequately covered in the existing data. However, obtaining, selecting, and labeling relevant data is expensive. This work presents a simple technique that automatically identifies new helpful examples suitable for human annotation. Our experimental results show that the proposed method, compared with random-selection and entropy-based methods, leads to higher accuracy improvements given a fixed annotation budget. Although developed and tested in the setting of a commercial intelligent assistant, the technique is of wider applicability.

Understanding the morphological changes of primary neuronal cells induced by chemical compounds is essential for drug discovery. Using the data from a single high-throughput imaging assay, a classification model for predicting the biological activity of candidate compounds was introduced. The image recognition model which is based on deep convolutional neural network (CNN) architecture with residual connections achieved accuracy of 99.6$\%$ on a binary classification task of distinguishing untreated and treated rodent primary neuronal cells with Amyloid-$\beta_{(25-35)}$.

An exciting new development in differential privacy is the shuffled model, which makes it possible to circumvent the large error lower bounds that are typically incurred in the local model, while relying on much weaker trust assumptions than in the central model. In this work, we study two basic statistical problems, namely, heavy hitters and $d$-dimensional range counting queries, in the shuffled model of privacy. For both problems we devise algorithms with polylogarithmic communication per user and polylogarithmic error; a consequence is an algorithm for approximating the median with similar communication and error. These bounds significantly improve on what is possible in the local model of differential privacy, where the error must provably grow polynomially with the number of users.

We study the problem of {\em properly} learning large margin halfspaces in the agnostic PAC model. In more detail, we study the complexity of properly learning $d$-dimensional halfspaces on the unit ball within misclassification error $\alpha \cdot \mathrm{OPT}_{\gamma} + \epsilon$, where $\mathrm{OPT}_{\gamma}$ is the optimal $\gamma$-margin error rate and $\alpha \geq 1$ is the approximation ratio. We give learning algorithms and computational hardness results for this problem, for all values of the approximation ratio $\alpha \geq 1$, that are nearly-matching for a range of parameters. Specifically, for the natural setting that $\alpha$ is any constant bigger than one, we provide an essentially tight complexity characterization. On the positive side, we give an $\alpha = 1.01$-approximate proper learner that uses $O(1/(\epsilon^2\gamma^2))$ samples (which is optimal) and runs in time $\mathrm{poly}(d/\epsilon) \cdot 2^{\tilde{O}(1/\gamma^2)}$. On the negative side, we show that {\em any} constant factor approximate proper learner has runtime $\mathrm{poly}(d/\epsilon) \cdot 2^{(1/\gamma)^{2-o(1)}}$, assuming the Exponential Time Hypothesis.

This paper presents an unsupervised method that trains neural source separation by using only multichannel mixture signals. Conventional neural separation methods require a lot of supervised data to achieve excellent performance. Although multichannel methods based on spatial information can work without such training data, they are often sensitive to parameter initialization and degraded with the sources located close to each other. The proposed method uses a cost function based on a spatial model called a complex Gaussian mixture model (cGMM). This model has the time-frequency (TF) masks and direction of arrivals (DoAs) of sources as latent variables and is used for training separation and localization networks that respectively estimate these variables. This joint training solves the frequency permutation ambiguity of the spatial model in a unified deep Bayesian framework. In addition, the pre-trained network can be used not only for conducting monaural separation but also for efficiently initializing a multichannel separation algorithm. Experimental results with simulated speech mixtures showed that our method outperformed a conventional initialization method.

Recent results in nonparametric regression show that deep learning, i.e., neural networks estimates with many hidden layers, are able to circumvent the so-called curse of dimensionality in case that suitable restrictions on the structure of the regression function hold. One key feature of the neural networks used in these results is that they are not fully connected. In this paper we show that we can get similar results also for fully connected multilayer feedforward neural networks with ReLU activation functions, provided the number of neurons per hidden layer is fixed and the number of hidden layers tends to infinity for sample size tending to infinity. The proof is based on new approximation results concerning fully connected deep neural networks.

In this paper, we settle the sampling complexity of solving discounted two-player turn-based zero-sum stochastic games up to polylogarithmic factors. Given a stochastic game with discount factor $\gamma\in(0,1)$ we provide an algorithm that computes an $\epsilon$-optimal strategy with high-probability given $\tilde{O}((1 - \gamma)^{-3} \epsilon^{-2})$ samples from the transition function for each state-action-pair. Our algorithm runs in time nearly linear in the number of samples and uses space nearly linear in the number of state-action pairs. As stochastic games generalize Markov decision processes (MDPs) our runtime and sample complexities are optimal due to Azar et al (2013). We achieve our results by showing how to generalize a near-optimal Q-learning based algorithms for MDP, in particular Sidford et al (2018), to two-player strategy computation algorithms. This overcomes limitations of standard Q-learning and strategy iteration or alternating minimization based approaches and we hope will pave the way for future reinforcement learning results by facilitating the extension of MDP results to multi-agent settings with little loss.

An e xtreme wind speed estimation method that consider s wind hazard climate type s is critical for design wind load calculation for building structure s affected by mixed climate s . However, it is very difficult to obtain wind hazard climate type s from meteorologi cal data records, because they restrict the application of extreme wind speed estimation in mixed climates . This paper first proposes a wind hazard type identification algorithm based on a numerical pattern recognition method that utilizes feature extraction and generalization . Next, it compares six commonly used machine learning models using K - fold cross - validation. Finally, it takes meteorological data from three locations near the southeast coast of China as example s to examine t he algor ithm's performance . Based on classification results, the extreme wind speed s calculated based on mixed wind hazard types is compared with those obtained from conventional methods, and the effects on structural design for different return periods are discus sed . Extreme wind speed; Mixed climates; Data - driven method; Pattern Recognition; Machine Learning; 1. Introduction Wind effects are key factors in structural design, and extreme wind speeds are the starting point . F or flexible structures such as long - span bridges, long - span roofs and high - rise buildings, wind loads are normally the predominant loads. I n order to meet both the ultimate safety and performance requirements of wind - resistant structural design, it s necessary to accurately estimate the extreme wind speed s for different recurrence period s . For significant buildings and infrastructures, it is necessary to estimat e the extreme wind speed through probabilistic methods from local wind speed record s .

In this paper, we propose a design of a model-free networked controller for a nonlinear plant whose mathematical model is unknown. In a networked control system, the controller and plant are located away from each other and exchange data over a network, which causes network delays that may fluctuate randomly due to network routing. So, in this paper, we assume that the current network delay is not known but the maximum value of fluctuating network delays is known beforehand. Moreover, we also assume that the sensor cannot observe all state variables of the plant. Under these assumption, we apply continuous deep Q-learning to the design of the networked controller. Then, we introduce an extended state consisting of a sequence of past control inputs and outputs as inputs to the deep neural network. By simulation, it is shown that, using the extended state, the controller can learn a control policy robust to the fluctuation of the network delays under the partial observation.