There has been an increased need for secondary means of credit evaluation by both traditional banking organizations as well as peer-to-peer lending entities. This is especially important in the present technological era where sticking with strict primary credit histories doesn't help distinguish between a 'good' and a 'bad' borrower, and ends up hurting both the individual borrower as well as the investor as a whole. We utilized machine learning classification and clustering algorithms to accurately predict a borrower's creditworthiness while identifying specific secondary attributes that contribute to this score. While extensive research has been done in predicting when a loan would be fully paid, the area of feature selection for lending is relatively new. We achieved 65% F1 and 73% AUC on the LendingClub data while identifying key secondary attributes.

Travel time estimation is an important component in modern transportation applications. The state of the art techniques for travel time estimation use GPS traces to learn the weights of a road network, often modeled as a directed graph, then apply Dijkstra-like algorithms to find shortest paths. Travel time is then computed as the sum of edge weights on the returned path. In order to enable time-dependency, existing systems compute multiple weighted graphs corresponding to different time windows. These graphs are often optimized offline before they are deployed into production routing engines, causing a serious engineering overhead. In this paper, we present STAD, a system that adjusts - on the fly - travel time estimates for any trip request expressed in the form of origin, destination, and departure time. STAD uses machine learning and sparse trips data to learn the imperfections of any basic routing engine, before it turns it into a full-fledged time-dependent system capable of adjusting travel times to real traffic conditions in a city. STAD leverages the spatio-temporal properties of traffic by combining spatial features such as departing and destination geographic zones with temporal features such as departing time and day to significantly improve the travel time estimates of the basic routing engine. Experiments on real trip datasets from Doha, New York City, and Porto show a reduction in median absolute errors of 14% in the first two cities and 29% in the latter. We also show that STAD performs better than different commercial and research baselines in all three cities.

This article characterizes the exact asymptotics of random Fourier feature (RFF) regression, in the realistic setting where the number of data samples $n$, their dimension $p$, and the dimension of feature space $N$ are all large and comparable. In this regime, the random RFF Gram matrix no longer converges to the well-known limiting Gaussian kernel matrix (as it does when $N \to \infty$ alone), but it still has a tractable behavior that is captured by our analysis. This analysis also provides accurate estimates of training and test regression errors for large $n,p,N$. Based on these estimates, a precise characterization of two qualitatively different phases of learning, including the phase transition between them, is provided; and the corresponding double descent test error curve is derived from this phase transition behavior. These results do not depend on strong assumptions on the data distribution, and they perfectly match empirical results on finite-dimensional real-world data sets.

We present an approach for unsupervised domain adaptation---with a strong focus on practical considerations of within-domain class imbalance and between-domain class distribution shift---from a class-conditioned domain alignment perspective. Current methods for class-conditioned domain alignment aim to explicitly minimize a loss function based on pseudo-label estimations of the target domain. However, these methods suffer from pseudo-label bias in the form of error accumulation. We propose a method that removes the need for explicit optimization of model parameters from pseudo-labels directly. Instead, we present a sampling-based implicit alignment approach, where the sample selection procedure is implicitly guided by the pseudo-labels. Theoretical analysis reveals the existence of a domain-discriminator shortcut in misaligned classes, which is addressed by the proposed implicit alignment approach to facilitate domain-adversarial learning. Empirical results and ablation studies confirm the effectiveness of the proposed approach, especially in the presence of within-domain class imbalance and between-domain class distribution shift.

Since collecting pixel-level groundtruth data is expensive, unsupervised visual understanding problems are currently an active research topic. In particular, several recent methods based on generative models have achieved promising results for object segmentation and saliency detection. However, since generative models are known to be unstable and sensitive to hyperparameters, the training of these methods can be challenging and time-consuming. In this work, we introduce an alternative, much simpler way to exploit generative models for unsupervised object segmentation. First, we explore the latent space of the BigBiGAN -- the state-of-the-art unsupervised GAN, which parameters are publicly available. We demonstrate that object saliency masks for GAN-produced images can be obtained automatically with BigBiGAN. These masks then are used to train a discriminative segmentation model. Being very simple and easy-to-reproduce, our approach provides competitive performance on common benchmarks in the unsupervised scenario.

Neural network pruning is an important technique for creating efficient machine learning models that can run on edge devices. We propose a new, highly flexible approach to neural network pruning based on Gibbs distributions. We apply it with Hamiltonians that are based on weight magnitude, using the annealing capabilities of Gibbs distributions to smoothly move from regularization to adaptive pruning during an ordinary neural network training schedule. This method can be used for either unstructured or structured pruning, and we provide explicit formulations for both. We compare our proposed method to several established pruning methods on ResNet variants and find that it outperforms them for unstructured, kernel-wise, and filter-wise pruning.

We consider incorporating incomplete physics knowledge, expressed as differential equations with latent functions, into Gaussian processes (GPs) to improve their performance, especially for limited data and extrapolation. While existing works have successfully encoded such knowledge via kernel convolution, they only apply to linear equations with analytical Green's functions. The convolution can further restrict us from fusing physics with highly expressive kernels, e.g., deep kernels. To overcome these limitations, we propose Physics Regularized Gaussian Process (PRGP) that can incorporate both linear and nonlinear equations, does not rely on Green's functions, and is free to use arbitrary kernels. Specifically, we integrate the standard GP with a generative model to encode the differential equation in a principled Bayesian hybrid framework. For efficient and effective inference, we marginalize out the latent variables and derive a simplified model evidence lower bound (ELBO), based on which we develop a stochastic collapsed inference algorithm. Our ELBO can be viewed as a posterior regularization objective. We show the advantage of our approach in both simulation and real-world applications.

The key task of physical simulation is to solve partial differential equations (PDEs) on discretized domains, which is known to be costly. In particular, high-fidelity solutions are much more expensive than low-fidelity ones. To reduce the cost, we consider novel Gaussian process (GP) models that leverage simulation examples of different fidelities to predict high-dimensional PDE solution outputs. Existing GP methods are either not scalable to high-dimensional outputs or lack effective strategies to integrate multi-fidelity examples. To address these issues, we propose Multi-Fidelity High-Order Gaussian Process (MFHoGP) that can capture complex correlations both between the outputs and between the fidelities to enhance solution estimation, and scale to large numbers of outputs. Based on a novel nonlinear coregionalization model, MFHoGP propagates bases throughout fidelities to fuse information, and places a deep matrix GP prior over the basis weights to capture the (nonlinear) relationships across the fidelities. To improve inference efficiency and quality, we use bases decomposition to largely reduce the model parameters, and layer-wise matrix Gaussian posteriors to capture the posterior dependency and to simplify the computation. Our stochastic variational learning algorithm successfully handles millions of outputs without extra sparse approximations. We show the advantages of our method in several typical applications.

Deep metric learning has gained promising improvement in recent years following the success of deep learning. It has been successfully applied to problems in few-shot learning, image retrieval, and open-set classifications. However, measuring the confidence of a deep metric learning model and identifying unreliable predictions is still an open challenge. This paper focuses on defining a calibrated and interpretable confidence metric that closely reflects its classification accuracy. While performing similarity comparison directly in the latent space using the learned distance metric, our approach approximates the distribution of data points for each class using a Gaussian kernel smoothing function. The post-processing calibration algorithm with proposed confidence metric on the held-out validation dataset improves generalization and robustness of state-of-the-art deep metric learning models while provides an interpretable estimation of the confidence. Extensive tests on four popular benchmark datasets (Caltech-UCSD Birds, Stanford Online Product, Stanford Car-196, and In-shop Clothes Retrieval) show consistent improvements even at the presence of distribution shifts in test data related to additional noise or adversarial examples.

This paper tries to present a more unified view of clustering, by identifying the relationships between five different clustering algorithms. Some of the results are not new, but they are presented in a cleaner, simpler and more concise way. To the best of my knowledge, the interpretation of DBSCAN as a climbing procedure, which introduces a theoretical connection between DBSCAN and Mean shift, is a novel result.