The standard objective in machine learning is to train a single model for all users. However, in many learning scenarios, such as cloud computing and federated learning, it is possible to learn one personalized model per user. In this work, we present a systematic learning-theoretic study of personalization. We propose and analyze three approaches: user clustering, data interpolation, and model interpolation. For all three approaches, we provide learning-theoretic guarantees and efficient algorithms for which we also demonstrate the performance empirically. All of our algorithms are model agnostic and work for any hypothesis class.

We study the linear subspace fitting problem in the overparameterized setting, where the estimated subspace can perfectly interpolate the training examples. Our scope includes the least-squares solutions to subspace fitting tasks with varying levels of supervision in the training data (i.e., the proportion of input-output examples of the desired low-dimensional mapping) and orthonormality of the vectors defining the learned operator. This flexible family of problems connects standard, unsupervised subspace fitting that enforces strict orthonormality with a corresponding regression task that is fully supervised and does not constrain the linear operator structure. This class of problems is defined over a supervision-orthonormality plane, where each coordinate induces a problem instance with a unique pair of supervision level and softness of orthonormality constraints. We explore this plane and show that the generalization errors of the corresponding subspace fitting problems follow double descent trends as the settings become more supervised and less orthonormally constrained.

Nowadays, devices are equipped with advanced sensors with higher processing/computing capabilities. Further, widespread Internet availability enables communication among sensing devices. As a result, vast amounts of data are generated on edge devices to drive Internet-of-Things (IoT), crowdsourcing, and other emerging technologies. The collected extensive data can be pre-processed, scaled, classified, and finally, used for predicting future events using machine learning (ML) methods. In traditional ML approaches, data is sent to and processed in a central server, which encounters communication overhead, processing delay, privacy leakage, and security issues. To overcome these challenges, each client can be trained locally based on its available data and by learning from the global model. This decentralized learning structure is referred to as Federated Learning (FL). However, in large-scale networks, there may be clients with varying computational resource capabilities. This may lead to implementation and scalability challenges for FL techniques. In this paper, we first introduce some recently implemented real-life applications of FL. We then emphasize on the core challenges of implementing the FL algorithms from the perspective of resource limitations (e.g., memory, bandwidth, and energy budget) of client clients. We finally discuss open issues associated with FL and highlight future directions in the FL area concerning resource-constrained devices.

We propose a statistical adaptive procedure called SALSA for automatically scheduling the learning rate (step size) in stochastic gradient methods. SALSA first uses a smoothed stochastic line-search procedure to gradually increase the learning rate, then automatically switches to a statistical method to decrease the learning rate. The line search procedure ``warms up'' the optimization process, reducing the need for expensive trial and error in setting an initial learning rate. The method for decreasing the learning rate is based on a new statistical test for detecting stationarity when using a constant step size. Unlike in prior work, our test applies to a broad class of stochastic gradient algorithms without modification. The combined method is highly robust and autonomous, and it matches the performance of the best hand-tuned learning rate schedules in our experiments on several deep learning tasks.

Stochastic gradient descent (SGD) with constant momentum and its variants such as Adam are the optimization algorithms of choice for training deep neural networks (DNNs). Since DNN training is incredibly computationally expensive, there is great interest in speeding up convergence. Nesterov accelerated gradient (NAG) improves the convergence rate of gradient descent (GD) for convex optimization using a specially designed momentum; however, it accumulates error when an inexact gradient is used (such as in SGD), slowing convergence at best and diverging at worst. In this paper, we propose Scheduled Restart SGD (SRSGD), a new NAG-style scheme for training DNNs. SRSGD replaces the constant momentum in SGD by the increasing momentum in NAG but stabilizes the iterations by resetting the momentum to zero according to a schedule. Using a variety of models and benchmarks for image classification, we demonstrate that, in training DNNs, SRSGD significantly improves convergence and generalization; for instance in training ResNet200 for ImageNet classification, SRSGD achieves an error rate of 20.93% vs. the benchmark of 22.13%. These improvements become more significant as the network grows deeper. Furthermore, on both CIFAR and ImageNet, SRSGD reaches similar or even better error rates with fewer training epochs compared to the SGD baseline. We provide code for SRSGD at https://github.com/minhtannguyen/SRSGD.

Given the knowledge of the separable structure of the function, one can design a sparse network to represent the function very accurately, or even exactly. When such structural information is not available, and we may only use a dense neural network, the optimization procedure to find the sparse network embedded in the dense network is similar to finding the needle in a haystack, using a given number of samples of the function. We demonstrate that the cost (measured by sample complexity) of finding the needle is directly related to the Barron norm of the function. While only a small number of samples is needed to train a sparse network, the dense network trained with the same number of samples exhibits large test loss and a large generalization gap. In order to control the size of the generalization gap, we find that the use of explicit regularization becomes increasingly more important as d increases. The numerically observed sample complexity with explicit regularization scales as O(d 2.5), which is in fact better than the theoretically predicted sample complexity that scales as O(d 4). Without explicit regularization (also called implicit regularization), the numerically observed sample complexity is significantly higher and is close to O(d 4.5).

We develop exact representations of two layer neural networks with rectified linear units in terms of a single convex program with number of variables polynomial in the number of training samples and number of hidden neurons. Our theory utilizes semi-infinite duality and minimum norm regularization. Moreover, we show that certain standard multi-layer convolutional neural networks are equivalent to L1 regularized linear models in a polynomial sized discrete Fourier feature space. We also introduce exact semi-definite programming representations of convolutional and fully connected linear multi-layer networks which are polynomial size in both the sample size and dimension.

Learning rich representation from data is an important task for deep generative models such as variational auto-encoder (VAE). However, by extracting high-level abstractions in the bottom-up inference process, the goal of preserving all factors of variations for top-down generation is compromised. Motivated by the concept of "starting small", we present a strategy to progressively learn independent hierarchical representations from high- to low-levels of abstractions. The model starts with learning the most abstract representation, and then progressively grow the network architecture to introduce new representations at different levels of abstraction. We quantitatively demonstrate the ability of the presented model to improve disentanglement in comparison to existing works on two benchmark data sets using three disentanglement metrics, including a new metric we proposed to complement the previously-presented metric of mutual information gap. We further present both qualitative and quantitative evidence on how the progression of learning improves disentangling of hierarchical representations. By drawing on the respective advantage of hierarchical representation learning and progressive learning, this is to our knowledge the first attempt to improve disentanglement by progressively growing the capacity of VAE to learn hierarchical representations.

We propose to compute Wasserstein barycenters (WBs) by solving for Monge maps with variational principle. We discuss the metric properties of WBs and explore their connections, especially the connections of Monge WBs, to K-means clustering and co-clustering. We also discuss the feasibility of Monge WBs on unbalanced measures and spherical domains. We propose two new problems -- regularized K-means and Wasserstein barycenter compression. We demonstrate the use of VWBs in solving these clustering-related problems.

We propose a stochastic variant of the classical Polyak step-size (Polyak, 1987) commonly used in the subgradient method. Although computing the Polyak step-size requires knowledge of the optimal function values, this information is readily available for typical modern machine learning applications. Consequently, the proposed stochastic Polyak step-size (SPS) is an attractive choice for setting the learning rate for stochastic gradient descent (SGD). We provide theoretical convergence guarantees for SGD equipped with SPS in different settings, including strongly convex, convex and non-convex functions. Furthermore, our analysis results in novel convergence guarantees for SGD with a constant step-size. We show that SPS is particularly effective when training over-parameterized models capable of interpolating the training data. In this setting, we prove that SPS enables SGD to converge to the true solution at a fast rate without requiring the knowledge of any problem-dependent constants or additional computational overhead. We experimentally validate our theoretical results via extensive experiments on synthetic and real datasets. We demonstrate the strong performance of SGD with SPS compared to state-of-the-art optimization methods when training over-parameterized models.