Coupled Matrix Tensor Factorization (CMTF) facilitates the integration and analysis of multiple data sources and helps discover meaningful information. Nonnegative CMTF (N-CMTF) has been employed in many applications for identifying latent patterns, prediction, and recommendation. However, due to the added complexity with coupling between tensor and matrix data, existing N-CMTF algorithms exhibit poor computation efficiency. In this paper, a computationally efficient N-CMTF factorization algorithm is presented based on the column-wise element selection, preventing frequent gradient updates. Theoretical and empirical analyses show that the proposed N-CMTF factorization algorithm is not only more accurate but also more computationally efficient than existing algorithms in approximating the tensor as well as in identifying the underlying nature of factors.

Simulated annealing is an effective and general means of optimization. It is in fact inspired by metallurgy, where the temperature of a material determines its behavior in thermodynamics. Likewise, in simulated annealing, the actions that the algorithm takes depend entirely on the value of a variable which captures the notion of temperature. Typically, simulated annealing starts with a high temperature, which makes the algorithm pretty unpredictable, and gradually cools the temperature down to become more stable. A key component that plays a crucial role in the performance of simulated annealing is the criteria under which the temperature changes namely, the cooling schedule. Motivated by this, we study the following question in this work: "Given enough samples to the instances of a specific class of optimization problems, can we design optimal (or approximately optimal) cooling schedules that minimize the runtime or maximize the success rate of the algorithm on average when the underlying problem is drawn uniformly at random from the same class?" We provide positive results both in terms of sample complexity and simulation complexity. For sample complexity, we show that $\tilde O(\sqrt{m})$ samples suffice to find an approximately optimal cooling schedule of length $m$. We complement this result by giving a lower bound of $\tilde \Omega(m^{1/3})$ on the sample complexity of any learning algorithm that provides an almost optimal cooling schedule. These results are general and rely on no assumption. For simulation complexity, however, we make additional assumptions to measure the success rate of an algorithm. To this end, we introduce the monotone stationary graph that models the performance of simulated annealing. Based on this model, we present polynomial time algorithms with provable guarantees for the learning problem.

Probabilistic programs with mixed support (both continuous and discrete latent random variables) commonly appear in many probabilistic programming systems (PPSs). However, the existence of the discrete random variables prohibits many basic gradient-based inference engines, which makes the inference procedure on such models particularly challenging. Existing PPSs either require the user to manually marginalize out the discrete variables or to perform a composing inference by running inference separately on discrete and continuous variables. The former is infeasible in most cases whereas the latter has some fundamental shortcomings. We present a novel approach to run inference efficiently and robustly in such programs using stochastic gradient Markov Chain Monte Carlo family of algorithms. We compare our stochastic gradient-based inference algorithm against conventional baselines in several important cases of probabilistic programs with mixed support, and demonstrate that it outperforms existing composing inference baselines and works almost as well as inference in marginalized versions of the programs, but with less programming effort and at a lower computation cost.

In this paper, we study the dynamics of gradient descent in learning neural networks for classification problems. Unlike in existing works, we consider the linearly non-separable case where the training data of different classes lie in orthogonal subspaces. We show that when the network has sufficient (but not exceedingly large) number of neurons, (1) the corresponding minimization problem has a desirable landscape where all critical points are global minima with perfect classification; (2) gradient descent is guaranteed to converge to the global minima in this case. Moreover, we discovered a geometric condition on the network weights so that when it is satisfied, the weight evolution transitions from a slow phase of weight direction spreading to a fast phase of weight convergence. The geometric condition says that the convex hull of the weights projected on the unit sphere contains the origin.

Withand without-replacement sampling of the individual component functions are regarded as some of the most popular variants of SGD. During SGD with replacement sampling, the stochastic gradient is equal to g(x, ξ i) f ξi (x) and ξ i is a uniform number in {1,..., n}, i.e., a with-replacement sample from the set of gradients f 1,..., f n . In the case of without-replacement sapling, the stochastic gradient is equal to g(x, ξ i) f ξi (x) and ξ i is the i-th ordered element in a random permutation of the numbers in {1,..., n}, i.e., a without-replacement sample. In practice, SGD without replacement is much more widely used compared to its with-replacement counterpart, as it can empirically converge significantly faster [1, 2, 3]. However, in the land of theoretical guarantees, with-replacement SGD has been the focal point of convergence analyses. The reason for this is that analyzing stochastic gradients born with replacement is significantly more tractable for a simple reason: in expectation, the stochastic gradient is equal to the "true" gradient of F, i.e., E ξi f ξi (x) F (x). This makes SGD amenable to analyses very similar to that of vanilla gradient descent (GD), which has been extensively studied under a large variety of function classes and geometric assumptions, e.g., see [4]. Unfortunately, the same cannot be said for SGD without replacement, which has long resisted nonvacuous convergence guaranteess.

Stochastic optimization algorithms, such as Stochastic Gradient Descent (SGD) and its variants, are mainstream methods for training deep networks in practice. However, the theoretical mechanism behind gradient noise still remains to be further investigated. Deep learning is known to find flat minima with a large neighboring region in parameter space from which each weight vector has similar small error. In this paper, we focus on a fundamental problem in deep learning, "How can deep learning usually find flat minima among so many minima?" To answer the question, we develop a density diffusion theory (DDT) for revealing the fundamental dynamical mechanism of SGD and deep learning. More specifically, we study how escape time from loss valleys to the outside of valleys depends on minima sharpness, gradient noise and hyperparameters. One of the most interesting findings is that stochastic gradient noise from SGD can help escape from sharp minima exponentially faster than flat minima, while white noise can only help escape from sharp minima polynomially faster than flat minima. We also find large-batch training requires exponentially many iterations to pass through sharp minima and find flat minima. We present direct empirical evidence supporting the proposed theoretical results.

Nowadays, privacy preserving machine learning has been drawing much attention in both industry and academy. Meanwhile, recommender systems have been extensively adopted by many commercial platforms (e.g. Amazon) and they are mainly built based on user-item interactions. Besides, social platforms (e.g. Facebook) have rich resources of user social information. It is well known that social information, which is rich on social platforms such as Facebook, are useful to recommender systems. It is anticipated to combine the social information with the user-item ratings to improve the overall recommendation performance. Most existing recommendation models are built based on the assumptions that the social information are available. However, different platforms are usually reluctant to (or cannot) share their data due to certain concerns. In this paper, we first propose a SEcure SOcial RECommendation (SeSoRec) framework which can (1) collaboratively mine knowledge from social platform to improve the recommendation performance of the rating platform, and (2) securely keep the raw data of both platforms. We then propose a Secret Sharing based Matrix Multiplication (SSMM) protocol to optimize SeSoRec and prove its correctness and security theoretically. By applying minibatch gradient descent, SeSoRec has linear time complexities in terms of both computation and communication. The comprehensive experimental results on three real-world datasets demonstrate the effectiveness of our proposed SeSoRec and SSMM.

In this paper, we introduce a powerful and efficient framework for the direct optimization of ranking metrics. The problem is ill-posed due to the discrete structure of the loss, and to deal with that, we introduce two important techniques: a stochastic smoothing and a novel gradient estimate based on partial integration. We also address the problem of smoothing bias and present a universal solution for a proper debiasing. To guarantee the global convergence of our method, we adopt a recently proposed Stochastic Gradient Langevin Boosting algorithm. Our algorithm is implemented as a part of the CatBoost gradient boosting library and outperforms the existing approaches on several learning to rank datasets. In addition to ranking metrics, our framework applies to any scale-free discreet loss function.

Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training and generalization behavior of infinitely wide two-layer neural networks with homogeneous activations. We show that the limits of the gradient flow on exponentially tailed losses can be fully characterized as a max-margin classifier in a certain non-Hilbertian space of functions. In presence of hidden low-dimensional structures, the resulting margin is independent of the ambiant dimension, which leads to strong generalization bounds. In contrast, training only the output layer implicitly solves a kernel support vector machine, which a priori does not enjoy such an adaptivity. Our analysis of training is non-quantitative in terms of running time but we prove computational guarantees in simplified settings by showing equivalences with online mirror descent. Finally, numerical experiments suggest that our analysis describes well the practical behavior of two-layer neural networks with ReLU activation and confirm the statistical benefits of this implicit bias.

Recent empirical works show that large deep neural networks are often highly redundant and one can find much smaller subnetworks without a significant drop of accuracy. However, most existing methods of network pruning are empirical and heuristic, leaving it open whether good subnetworks provably exist, how to find them efficiently, and if network pruning can be provably better than direct training using gradient descent. We answer these problems positively by proposing a simple greedy selection approach for finding good subnetworks, which starts from an empty network and greedily adds important neurons from the large network. This differs from the existing methods based on backward elimination, which remove redundant neurons from the large network. Theoretically, applying our greedy selection strategy on sufficiently large pre-trained networks guarantees to find small subnetworks with lower loss than networks directly trained with gradient descent. Practically, we improve prior arts of network pruning on learning compact neural architectures on ImageNet, including ResNet, MobilenetV2/V3, and ProxylessNet. Our theory and empirical results on MobileNet suggest that we should fine-tune the pruned subnetworks to leverage the information from the large model, instead of re-training from new random initialization as suggested in \citet{liu2018rethinking}.