arXiv subjects

Emily Zhu

Publications and source records attributed to Emily Zhu.

7 recordsLinked to original sources

arXiv2025

Distributionally Robust Predictive Runtime Verification under Spatio-Temporal Logic Specifications

Cyber-physical systems (CPS) designed in simulators, often consisting of multiple interacting agents (e.g. in multi-agent formations), behave differently in the real-world. We want to verify these systems during runtime when they are deployed. We thus propose robust predictive runtime verification (RPRV) algorithms for: (1) general stochastic CPS under signal temporal logic (STL) tasks, and (2) stochastic multi-agent systems (MAS) under spatio-temporal logic tasks. The RPRV problem presents the following challenges: (1) there may not be sufficient data on the behavior of the deployed CPS, (2) predictive models based on design phase system trajectories may encounter distribution shift during real-world deployment, and (3) the algorithms need to scale to the complexity of MAS and be applicable to spatio-temporal logic tasks. To address the challenges, we assume knowledge of an upper bound on the statistical distance between the trajectory distributions of the system at deployment and design time. We are motivated by our prior work [1, 2] where we proposed an accurate and an interpretable RPRV algorithm for general CPS, which we here extend to the MAS setting and spatio-temporal logic tasks. Specifically, we use a learned predictive model to estimate the system behavior at runtime and robust conformal prediction to obtain probabilistic guarantees by accounting for distribution shifts. Building on [1], we perform robust conformal prediction over the robust semantics of spatio-temporal reach and escape logic (STREL) to obtain centralized RPRV algorithms for MAS. We empirically validate our results in a drone swarm simulator, where we show the scalability of our RPRV algorithms to MAS and analyze the impact of different trajectory predictors on the verification result. To the best of our knowledge, these are the first statistically valid algorithms for MAS under distribution shift.

eess.SY↗

arXiv2024

Two-Point Concentration of the Domination Number of Random Graphs

We show that the domination number of the binomial random graph G_{n,p} with edge-probability p is concentrated on two values for p \ge n^{-2/3+\eps}, and not concentrated on two values for general p \le n^{-2/3}. This refutes a conjecture of Glebov, Liebenau and Szabo, who showed two-point concentration for p \ge n^{-1/2+\eps}, and conjectured that two-point concentration fails for p \ll n^{-1/2}. The proof of our main result requires a Poisson type approximation for the probability that a random bipartite graph has no isolated vertices, in a regime where standard tools are unavailable (as the expected number of isolated vertices is relatively large). We achieve this approximation by adapting the proof of Janson's inequality to this situation, and this adaptation may be of broader interest.

math.PR↗

arXiv2023

Isomorphisms between dense random graphs

We consider two variants of the induced subgraph isomorphism problem for two independent binomial random graphs with constant edge-probabilities p_1,p_2. In particular, (i) we prove a sharp threshold result for the appearance of G_{n,p_1} as an induced subgraph of G_{N,p_2}, (ii) we show two-point concentration of the size of the maximum common induced subgraph of G_{N, p_1} and G_{N,p_2}, and (iii) we show that the number of induced copies of G_{n,p_1} in G_{N,p_2} has an unusual limiting distribution. These results confirm simulation-based predictions of McCreesh, Prosser, Solnon and Trimble, and resolve several open problems of Chatterjee and Diaconis. The proofs are based on careful refinements of the first and second moment method, using extra twists to (a) take some non-standard behaviors into account, and (b) work around the large variance issues that prevent standard applications of these methods.

math.CO↗

arXiv2023

On multicolor Ramsey numbers of triple system paths of length 3

Let $\mathcal{H}$ be a 3-uniform hypergraph. The multicolor Ramsey number $ r_k(\mathcal{H})$ is the smallest integer $n$ such that every coloring of $ \binom{[n]}{3}$ with $k$ colors has a monochromatic copy of $\mathcal{H}$. Let $ \mathcal{L}$ be the loose 3-uniform path with 3 edges and $ \mathcal{M}$ denote the messy 3-uniform path with 3 edges; that is, let $\mathcal{L} = \{abc, cde, efg\}$ and $\mathcal{M} = \{ abc, bcd, def\}$. In this note we prove $ r_k(\mathcal{L}) < 1.54k$ and $ r_k(\mathcal{M}) < 1.6k$ for $k$ sufficiently large.

math.CO↗

arXiv2020

A note on the Erdős-Hajnal hypergraph Ramsey problem

We show that there is an absolute constant $c>0$ such that the following holds. For every $n > 1$, there is a 5-uniform hypergraph on at least $2^{2^{cn^{1/4}}}$ vertices with independence number at most $n$, where every set of 6 vertices induces at most 3 edges. The double exponential growth rate for the number of vertices is sharp. By applying a stepping-up lemma established by the first two authors, analogous sharp results are proved for $k$-uniform hypergraphs. This answers the penultimate open case of a conjecture in Ramsey theory posed by Erdős and Hajnal in 1972.

math.CO↗

arXiv2018

Multicast Triangular Semilattice Network

We investigate the structure of the code graph of a multicast network that has a characteristic shape of an inverted equilateral triangle. We provide a criterion that determines the validity of a receiver placement within the code graph, present invariance properties of the determinants corresponding to receiver placements under symmetries, and provide a complete study of these networks' receivers and required field sizes up to a network of 4 sources. We also improve on various definitions related to code graphs.

cs.IT↗

arXiv2015

Modeling Galactic Conformity with the Color-Halo Age Relation in the Illustris Simulation

Comparisons between observational surveys and galaxy formation models find that the mass of dark matter haloes can largely explain galaxies' stellar mass. However, it remains uncertain whether additional environmental variables, generally referred to as assembly bias, are necessary to explain other galaxy properties. We use the Illustris Simulation to investigate the role of assembly bias in producing galactic conformity by considering 18,000 galaxies with $M_{stellar}$ > $2 \times 10^9$ $M_{\odot}$. We find a significant signal of galactic conformity: out to distances of about 10 Mpc, the mean red fraction of galaxies around redder galaxies is higher than around bluer galaxies at fixed stellar mass. Dark matter haloes exhibit an analogous conformity signal, in which the fraction of haloes formed at earlier times (old haloes) is higher around old haloes than around younger ones at fixed halo mass. A plausible interpretation of galactic conformity can be given as a combination of the halo conformity signal with the galaxy color-halo age relation: at fixed stellar mass, particularly toward the low-mass end, Illustris' galaxy colors correlate with halo age, with the reddest galaxies (often satellites) being preferentially found in the oldest haloes. In fact, we can explain the galactic conformity effect with a simple semi-empirical model, by assigning stellar mass based on halo mass (abundance matching) and by assigning galaxy color based on halo age (age matching). We investigate other interpretations for the galactic conformity, particularly its dependence on the isolation criterion and on the central-satellite information. Regarding comparison to observations, we conclude that the adopted selection/isolation criteria, projection effects, and stacking techniques can have a significant impact on the measured amplitude of the conformity signal.

astro-ph.GA↗