Laboratoire d’Informatique Fondamentale de Marseille

Most work on verbalising Knowledge-Graphs (KG) has focused on high-resource languages such as English, Russian, Czech or Arabic. In this paper, we focus on KG-to-Text generation where the output text is in Breton, Irish or Welsh. To overcome the small size of the parallel training data, we combine the strengths of a multilingual encoder-decoder model with denoising fine-tuning on monolingual data and Soft Prompt fine-tuning on a small quantity of KG/text data. We furthermore structure the soft prompt into multiple sub-prompts designed to capture the similarities and differences between English, Knowledge graphs and the three target languages. Our experiments show that our approach outperforms strong baselines and that all sub-prompts contribute to performance 1 .

We consider the Trivially Perfect Editing problem, where one is given an undirected graph $G = (V,E)$ and a parameter $k \in \mathbb{N}$ and seeks to edit (add or delete) at most $k$ edges from $G$ to obtain a trivially perfect graph. The related Trivially Perfect Completion and Trivially Perfect Deletion problems are obtained by only allowing edge additions or edge deletions, respectively. Trivially perfect graphs are both chordal and cographs, and have applications related to the tree-depth width parameter and to social network analysis. All variants of the problem are known to be NP-Complete and to admit so-called polynomial kernels. More precisely, the existence of an $O(k^3)$ vertex-kernel for Trivially Perfect Completion was announced by Guo (ISAAC 2007) but without a stand-alone proof. More recently, Drange and Pilipczuk (Algorithmica 2018) provided $O(k^7)$ vertex-kernels for these problems and left open the existence of cubic vertex-kernels. In this work, we answer positively to this question for all three variants of the problem.

Summary We show in this work that using the NSIMD vectorization library allowed to obtain better performances on the EFISPEC3D kernel, a spectral-finite-element method to solve the forward seismic wave propagation problem. Moreover the same code without modification can be compiled to target different SIMD extensions with different vector sizes without degrading performances.

Many algorithms designed for shared-memory distributed systems assume the single-writer multi- reader (SWMR) setting where each process is provided with a unique register that can only be written by the process and read by all. In a system where computation is performed by a bounded number n of processes coming from a large (possibly unbounded) set of potential participants, the assumption of an SWMR memory is no longer reasonable. If only a bounded number of multi- writer multi-reader (MWMR) registers are provided, we cannot rely on an a priori assignment of processes to registers. In this setting, implementing an SWMR memory, or equivalently, ensuring stable writes (i.e., every written value persists in the memory), is desirable. In this paper, we propose an SWMR implementation that adapts the number of MWMR registers used to the desired progress condition. For any given k from 1 to n, we present an algorithm that uses n + k − 1 registers to implement a k-lock-free SWMR memory. In the special case of 2-lock-freedom, we also give a matching lower bound of n + 1 registers, which supports our conjecture that the algorithm is space-optimal. Our lower bound holds for the strictly weaker progress condition of 2-obstruction-freedom, which suggests that the space complexity for k-obstruction-free and k-lock-free SWMR implementations might coincide.

Bulk Synchronous Parallel (BSP) is a model for parallel computing with predictable scalability. BSP has a cost model: programs can be assigned a cost which describes their resource usage on any parallel machine. However, the programmer has to manually derive this cost. This paper describes an automatic method for the derivation of BSP program costs, based on classic cost analysis and approximation of polyhedral integer volumes. Our method requires and analyzes programs with textually aligned synchronization and textually aligned, polyhedral communication. We have implemented the analysis and our prototype obtains cost formulas that are parametric in the input parameters of the program and the parameters of the BSP computer and thus bound the cost of running the program with any input on any number of cores. We evaluate the cost formulas and find that they are indeed upper bounds, and tight for data-oblivious programs. Additionally, we evaluate their capacity to predict concrete run times in two parallel settings: a multi-core computer and a cluster. We find that when exact upper bounds can be found, they accurately predict run-times. In networks with full bisection bandwidth, as the BSP model supposes, results are promising with errors < 50%.

Many reasoning problems are based on the problem of satisfiability (SAT). While SAT itself becomes easy when restricting the structure of the formulas in a certain way, the situation is more opaque for more involved decision problems. We consider here the CardMinSat problem which asks, given a propositional formula $\phi$ and an atom $x$, whether $x$ is true in some cardinality-minimal model of $\phi$. This problem is easy for the Horn fragment, but, as we will show in this paper, remains $\Theta_2$-complete (and thus $\mathrm{NP}$-hard) for the Krom fragment (which is given by formulas in CNF where clauses have at most two literals). We will make use of this fact to study the complexity of reasoning tasks in belief revision and logic-based abduction and show that, while in some cases the restriction to Krom formulas leads to a decrease of complexity, in others it does not. We thus also consider the CardMinSat problem with respect to additional restrictions to Krom formulas towards a better understanding of the tractability frontier of such problems.