arXiv research

A simple, flexible approach to creating expressive priors in Gaussian process (GP) models makes new kernels from a combination of basic kernels, e.g. summing a periodic and linear kernel can capture seasonal variation with a long term trend. Despite a well-studied link between GPs and Bayesian neural networks (BNNs), t…

Periodicity is often studied in timeseries modelling with autoregressive methods but is less popular in the kernel literature, particularly for higher dimensional problems such as in textures, crystallography, and quantum mechanics. Large datasets often make modelling periodicity untenable for otherwise powerful non-pa…

The paper proposes a method to improve Bayesian inference for periodic data using data-driven priors.

The Bergman kernel and period map for curves are studied.

The X-ray transform on the periodic slab $[0,1]\times\mathbb T^n$, $n\geq0$, has a non-trivial kernel due to the symmetry of the manifold and presence of trapped geodesics. For tensor fields gauge freedom increases the kernel further, and the X-ray transform is not solenoidally injective unless $n=0$. We characterize t…

Advanced kernels improve Gaussian process accuracy by incorporating domain knowledge.

We show that, when considering the anisotropic scaling factors and their derivatives as affine variables, the coefficients of the heat kernel expansion of the Dirac-Laplacian on $SU(2)$ Bianchi IX metrics are algebro-geometric periods of motives of complements in affine spaces of unions of quadrics and hyperplanes. We …

Study Bergman kernel metrics on degenerating hyperelliptic surfaces.

Develops a GP framework for age and year-specific mortality surfaces.

New method for optimizing risk in financial models using Fourier transforms.

The paper extends kernel ridge regression to product kernels and reveals new convergence behaviors.

GP-LSTM model predicts stock returns and volatility more accurately.

Generative model captures repetitive industrial processes with varying durations and dynamics.

We introduce the convolutional spectral kernel (CSK), a novel family of non-stationary, nonparametric covariance kernels for Gaussian process (GP) models, derived from the convolution between two imaginary radial basis functions. We present a principled framework to interpret CSK, as well as other deep probabilistic mo…

Knitted and woven textile structures are examples of doubly periodic structures in a thickened plane made out of intertwining strands of yarn. Factoring out the group of translation symmetries of such a structure gives rise to a link diagram in a thickened torus. Such a diagram on a standard torus is converted into a c…

Gaussian processes are rich distributions over functions, with generalization properties determined by a kernel function. When used for long-range extrapolation, predictions are particularly sensitive to the choice of kernel parameters. It is therefore critical to account for kernel uncertainty in our predictive distri…

We consider a parallelizable $2n$-manifold $F$ which has the homotopy type of the wedge product of $n$-spheres and show that the group of pseudo-isotopy classes of orientation preserving diffeomorphisms that keep the boundary $\partial F$ pointwise fixed and induce the trivial variation operator is a central extension …

We prove that the hyperelliptic Torelli group is generated by Dehn twists about separating curves that are preserved by the hyperelliptic involution. This verifies a conjecture of Hain. The hyperelliptic Torelli group can be identified with the kernel of the Burau representation evaluated at t=-1 and also the fundament…

Deformations of compact Riemann surfaces are considered using a Čech cohomology sliding overlaps approach. Cocycles are calculated for conformal cutting and regluing deformations at zeros of Abelian differentials. A second order deformation expansion is presented for the Riemann period matrix. A complete deformation ex…

Given an invariant gauge potential and a periodic scalar potential \tilde{V} on a Riemannian manifold \tilde{M} with a discrete symmetry group Γ, consider a Γ-periodic quantum Hamiltonian \tilde{H}=-\tildeΔ_{B}+\tilde{V} where \tildeΔ_{B} is the Bochner Laplacian. Both the gauge group and the symmetry group Γcan be non…

Predicting ambulance demand accurately at a fine resolution in time and space (e.g., every hour and 1 km$^2$) is critical for staff / fleet management and dynamic deployment. There are several challenges: though the dataset is typically large-scale, demand per time period and locality is almost always zero. The demand …

Study on optimal rate of kernel regression for large-dimensional data.

PROPO tackles non-stationary MDPs with efficient policy optimization.

One of the peculiarities of power and gas markets is the delivery mechanism of forward contracts. The seller of a futures contract commits to deliver, say, power, over a certain period, while the classical forward is a financial agreement settled on a maturity date. Our purpose is to design a Heath-Jarrow-Morton framew…

New GP kernels avoid mean reversion without losing smoothness.

Study shows solutions of differential inclusions are homotopy equivalent in $W^{1,p}$-topology.

Researchers develop a generalised geometric Brownian motion for better asset pricing.

Cellular network configuration plays a critical role in network performance. In current practice, network configuration depends heavily on field experience of engineers and often remains static for a long period of time. This practice is far from optimal. To address this limitation, online-learning-based approaches hav…

Let M be a complete Riemannian manifold with a free cocompact Z^k-action. Let k(t,x,y) be the heat kernel on M. We compute the asymptotics of k(t,x,y) in the limit in which t goes to infinity and d(x,y) is comparable to sqrt{t}. We show that in this limit, the heat diffusion is governed by an effective Euclidean metric…

Study small eigenvalues of Riemann surfaces degenerating with Kähler metrics.

Bayesian realized EGARCH models improve tail risk forecasting.

While frame-independent predictions with deep neural networks have become the prominent solutions to many computer vision tasks, the potential benefits of utilizing correlations between frames have received less attention. Even though probabilistic machine learning provides the ability to encode correlation as prior kn…

Computational efficiency is an important consideration for deploying machine learning models for time series prediction in an online setting. Machine learning algorithms adjust model parameters automatically based on the data, but often require users to set additional parameters, known as hyperparameters. Hyperparamete…

PIML uses physics equations in machine learning for better forecasting.

Let $Σ$ be a surface with either boundary or marked points, equipped with an arbitrary framing. In this note we determine the action of the associated "framed mapping class group" on the homology of $Σ$ relative to its boundary (respectively marked points), describing the image as the kernel of a certain crossed homomo…

In this paper, we propose a new threshold-kernel jump-detection method for jump-diffusion processes, which iteratively applies thresholding and kernel methods in an approximately optimal way to achieve improved finite-sample performance. We use the expected number of jump misclassifications as the objective function to…

Kernel-based Bayesian filter for nonlinear systems using infinite-dimensional operators.

Combines physics-based ML with hierarchical Bayesian techniques for better model performance.

Regularized Stein thinning improves MCMC output approximations.

We model the arrival of mid-price changes in the E-Mini S&P futures contract as a self-exciting Hawkes process. Using several estimation methods, we find that the Hawkes kernel is power-law with a decay exponent close to -1.15 at short times, less than approximately 10^3 seconds, and crosses over to a second power-law …

New method uses trainable activations to make BNNs behave like GPs.

Model inference for dynamical systems aims to estimate the future behaviour of a system from observations. Purely model-free statistical methods, such as Artificial Neural Networks, tend to perform poorly for such tasks. They are therefore not well suited to many questions from applications, for example in Bayesian fil…

GP model for time series forecasting with priors.

Social networks are getting closer to our real physical world. People share the exact location and time of their check-ins and are influenced by their friends. Modeling the spatio-temporal behavior of users in social networks is of great importance for predicting the future behavior of users, controlling the users' mov…

Probabilistic neural networks are typically modeled with independent weight priors, which do not capture weight correlations in the prior and do not provide a parsimonious interface to express properties in function space. A desirable class of priors would represent weights compactly, capture correlations between weigh…

Many fits of Hawkes processes to financial data look rather good but most of them are not statistically significant. This raises the question of what part of market dynamics this model is able to account for exactly. We document the accuracy of such processes as one varies the time interval of calibration and compare t…

In a pathbreaking paper, Cover and Ordentlich (1998) solved a max-min portfolio game between a trader (who picks an entire trading algorithm, $θ(\cdot)$) and "nature," who picks the matrix $X$ of gross-returns of all stocks in all periods. Their (zero-sum) game has the payoff kernel $W_θ(X)/D(X)$, where $W_θ(X)$ is the…

Machine-learning of atomic-scale properties amounts to extracting correlations between structure, composition and the quantity that one wants to predict. Representing the input structure in a way that best reflects such correlations makes it possible to improve the accuracy of the model for a given amount of reference …