Andrey Y. Lokhov

Inferring the origin of an epidemic with a dynamic message-passing algorithm.

We study the problem of estimating the origin of an epidemic outbreak: given a contact network and a snapshot of epidemic spread at a certain time, determine the infection source. This problem is important in different contexts of computer or social networks. Assuming that the epidemic spread follows the usual susceptible-infected-recovered model, we introduce an inference algorithm based on dynamic message-passing equations and we show that it leads to significant improvement of performance compared to existing approaches. Importantly, this algorithm remains efficient in the case where the snapshot sees only a part of the network.

Dynamic message-passing equations for models with unidirectional dynamics

Understanding and quantifying the dynamics of disordered out-of-equilibrium models is an important problem in many branches of science. Using the dynamic cavity method on time trajectories, we construct a general procedure for deriving the dynamic message-passing equations for a large class of models with unidirectional dynamics, which includes the zero-temperature random-field Ising model, the susceptible-infected-recovered model, and rumor spreading models. We show that unidirectionality of the dynamics is the key ingredient that makes the problem solvable. These equations are applicable to single instances of the corresponding problems with arbitrary initial conditions and are asymptotically exact for problems defined on locally treelike graphs. When applied to real-world networks, they generically provide a good analytic approximation of the real dynamics.

Optimal deployment of resources for maximizing impact in spreading processes

Significance Spreading processes play an increasingly important role in marketing, opinion setting, and epidemic modeling. Most existing algorithms for optimal resource allocation in spreading processes are based on topological characteristics of the underlying network and aim to maximize impact at infinite time. Clearly, realistic and efficient real-time allocation policies should consider both network properties and details of the dynamics; additionally, control may be applied only to a restricted set of accessible nodes, and impact should be maximized in a limited time window. We introduce a probabilistic targeting framework that incorporates the dynamics and encompasses previously considered optimization formulations. It is based on a scalable dynamic message-passing approach that allows for the solution of large real-world network instances. The effective use of limited resources for controlling spreading processes on networks is of prime significance in diverse contexts, ranging from the identification of “influential spreaders” for maximizing information dissemination and targeted interventions in regulatory networks, to the development of mitigation policies for infectious diseases and financial contagion in economic systems. Solutions for these optimization tasks that are based purely on topological arguments are not fully satisfactory; in realistic settings, the problem is often characterized by heterogeneous interactions and requires interventions in a dynamic fashion over a finite time window via a restricted set of controllable nodes. The optimal distribution of available resources hence results from an interplay between network topology and spreading dynamics. We show how these problems can be addressed as particular instances of a universal analytical framework based on a scalable dynamic message-passing approach and demonstrate the efficacy of the method on a variety of real-world examples.

Optimal structure and parameter learning of Ising models

An arbitrary Ising model can be exactly recovered from observations using an information-theoretically optimal amount of data. Reconstruction of the structure and parameters of an Ising model from binary samples is a problem of practical importance in a variety of disciplines, ranging from statistical physics and computational biology to image processing and machine learning. The focus of the research community shifted toward developing universal reconstruction algorithms that are both computationally efficient and require the minimal amount of expensive data. We introduce a new method, interaction screening, which accurately estimates model parameters using local optimization problems. The algorithm provably achieves perfect graph structure recovery with an information-theoretically optimal number of samples, notably in the low-temperature regime, which is known to be the hardest for learning. The efficacy of interaction screening is assessed through extensive numerical tests on synthetic Ising models of various topologies with different types of interactions, as well as on real data produced by a D-Wave quantum computer. This study shows that the interaction screening method is an exact, tractable, and optimal technique that universally solves the inverse Ising problem.

Detection of Cyber-Physical Faults and Intrusions from Physical Correlations

Cyber-physical systems are critical infrastructures that are crucial both to the reliable delivery of resources such as energy, and to the stable functioning of automatic and control architectures. These systems are composed of interdependent physical, control and communications networks described by disparate mathematical models creating scientific challenges that go well beyond the modeling and analysis of the individual networks. A key challenge in cyber-physical defense is a fast online detection and localization of faults and intrusions without prior knowledge of the failure type. We describe a set of techniques for the efficient identification of faults from correlations in physical signals, assuming only a minimal amount of available system information. The performance of our detection method is illustrated on data collected from a large building automation system.

Discovering a Transferable Charge Assignment Model Using Machine Learning

Partial atomic charge assignment is of immense practical value to force field parametrization, molecular docking, and cheminformatics. Machine learning has emerged as a powerful tool for modeling chemistry at unprecedented computational speeds given accurate reference data. However, certain tasks, such as charge assignment, do not have a unique solution. Herein, we use a machine learning algorithm to discover a new charge assignment model by learning to replicate molecular dipole moments across a large, diverse set of nonequilibrium conformations of molecules containing C, H, N, and O atoms. The new model, called Affordable Charge Assignment (ACA), is computationally inexpensive and predicts dipoles of out-of-sample molecules accurately. Furthermore, dipole-inferred ACA charges are transferable to dipole and even quadrupole moments of much larger molecules than those used for training. We apply ACA to dynamical trajectories of biomolecules and produce their infrared spectra. Additionally, we find that ACA assigns similar charges to Charge Model 5 but with greatly reduced computational cost.

Transferable Dynamic Molecular Charge Assignment Using Deep Neural Networks

The ability to accurately and efficiently compute quantum-mechanical partial atomistic charges has many practical applications, such as calculations of IR spectra, analysis of chemical bonding, and classical force field parametrization. Machine learning (ML) techniques provide a possible avenue for the efficient prediction of atomic partial charges. Modern ML advances in the prediction of molecular energies [i.e., the hierarchical interacting particle neural network (HIP-NN)] has provided the necessary model framework and architecture to predict transferable, extensible, and conformationally dynamic atomic partial charges based on reference density functional theory (DFT) simulations. Utilizing HIP-NN, we show that ML charge prediction can be highly accurate over a wide range of molecules (both small and large) across a variety of charge partitioning schemes such as the Hirshfeld, CM5, MSK, and NBO methods. To demonstrate transferability and size extensibility, we compare ML results with reference DFT calculations on the COMP6 benchmark, achieving errors of 0.004e- (elementary charge). This is remarkable since this benchmark contains two proteins that are multiple times larger than the largest molecules in the training set. An application of our atomic charge predictions on nonequilibrium geometries is the generation of IR spectra for organic molecules from dynamical trajectories on a variety of organic molecules, which show good agreement with calculated IR spectra with reference method. Critically, HIP-NN charge predictions are many orders of magnitude faster than direct DFT calculations. These combined results provide further evidence that ML (specifically HIP-NN) provides a pathway to greatly increase the range of feasible simulations while retaining quantum-level accuracy.

Topological transition in disordered planar matching: combinatorial arcs expansion

In this paper, we investigate analytically the properties of the disordered Bernoulli model of planar matching. This model is characterized by a topological phase transition, yielding complete planar matching solutions only above a critical density threshold. We develop a combinatorial procedure of arcs expansion that explicitly takes into account the contribution of short arcs, and allows to obtain an accurate analytical estimation of the critical value by reducing the global constrained problem to a set of local ones. As an application to a toy representation of the RNA secondary structures, we suggest generalized models that incorporate a one-to-one correspondence between the contact matrix and the RNA-type sequence, thus giving sense to the notion of effective non-integer alphabets.

Phase transition in random planar diagrams and RNA-type matching.

We study the planar matching problem, defined by a symmetric random matrix with independent identically distributed entries, taking values zero and one. We show that the existence of a perfect planar matching structure is possible only above a certain critical density, p(c), of allowed contacts (i.e., of ones). Using a formulation of the problem in terms of Dyck paths and a matrix model of planar contact structures, we provide an analytical estimation for the value of the transition point, p(c), in the thermodynamic limit. This estimation is close to the critical value, p(c)≈0.379, obtained in numerical simulations based on an exact dynamical programming algorithm. We characterize the corresponding critical behavior of the model and discuss the relation of the perfect-imperfect matching transition to the known molten-glass transition in the context of random RNA secondary structure formation. In particular, we provide strong evidence supporting the conjecture that the molten-glass transition at T=0 occurs at p(c).