Search

This paper presents the second-place methodology in the Volvo Discovery Challenge at ECML-PKDD 2024, where we used Long Short-Term Memory networks and pseudo-labeling to predict maintenance needs for a component of Volvo trucks. We processed the training data to mirror the test set structure and applied a base LSTM model to label the test data iteratively. This approach refined our model's predictive capabilities and culminated in a macro-average F1-score of 0.879, demonstrating robust performance in predictive maintenance. This work provides valuable insights for applying machine learning techniques effectively in industrial settings., Comment: 2nd place at ECML-PKDD Discovery Challenge https://www.hh.se/english/about-the-university/events/discovery-challenge-ecml-pkdd-2024.html

In 2021, Adam Zsolt Wagner proposed an approach to disprove conjectures in graph theory using Reinforcement Learning (RL). Wagner's idea can be framed as follows: consider a conjecture, such as a certain quantity f(G) < 0 for every graph G; one can then play a single-player graph-building game, where at each turn the player decides whether to add an edge or not. The game ends when all edges have been considered, resulting in a certain graph G_T, and f(G_T) is the final score of the game; RL is then used to maximize this score. This brilliant idea is as simple as innovative, and it lends itself to systematic generalization. Several different single-player graph-building games can be employed, along with various RL algorithms. Moreover, RL maximizes the cumulative reward, allowing for step-by-step rewards instead of a single final score, provided the final cumulative reward represents the quantity of interest f(G_T). In this paper, we discuss these and various other choices that can be significant in Wagner's framework. As a contribution to this systematization, we present four distinct single-player graph-building games. Each game employs both a step-by-step reward system and a single final score. We also propose a principled approach to select the most suitable neural network architecture for any given conjecture, and introduce a new dataset of graphs labeled with their Laplacian spectra. Furthermore, we provide a counterexample for a conjecture regarding the sum of the matching number and the spectral radius, which is simpler than the example provided in Wagner's original paper. The games have been implemented as environments in the Gymnasium framework, and along with the dataset, are available as open-source supplementary materials., Comment: Accepted at the 27th International Conference on Discovery Science http://ds2024.isti.cnr.it/

Cellulose nanocrystals (CNCs) are bioderived nanoparticles that can be isolated from various sources of natural cellulose via acid hydrolysis. However, the link between particle morphological characteristics and their ensemble behavior is poorly understood, partly because of the difficulties in controlling the CNC morphology during their extraction process. In this work, the impacts of common post-hydrolysis treatments on the CNC morphology are investigated. The results indicate that the centrifugation step commonly applied during CNC purification favors the formation of composite particles made of aligned crystallites, referred to as 'bundles'. Scanning nanobeam electron diffraction reveals that such bundles are associated preferentially along their hydrophobic faces. This is in stark contrast to the formation of misaligned composite particles that can be achieved with ionic treatments, where an uncontrolled aggregation occurs. The functional relevance of these morphological differences is demonstrated by their effect on the cholesteric self-organization of CNCs, with bundles found to exhibit a greater chiral enhancement, whereas the misaligned composite particles found to promote gelation. This study reveals the importance of the often-disregarded purification steps on the final morphology of CNCs and their resulting ensemble properties, thereby unlocking new routes for tailoring this promising material towards a variety of applications.

Community detection for time series without prior knowledge poses an open challenge within complex networks theory. Traditional approaches begin by assessing time series correlations and maximizing modularity under diverse null models. These methods suffer from assuming temporal stationarity and are influenced by the granularity of observation intervals. In this study, we propose an approach based on the signature matrix, a concept from path theory for studying stochastic processes. By employing a signature-derived similarity measure, our method overcomes drawbacks of traditional correlation-based techniques. Through a series of numerical experiments, we demonstrate that our method consistently yields higher modularity compared to baseline models, when tested on the Standard and Poor's 500 dataset. Moreover, our approach showcases enhanced stability in modularity when the length of the underlying time series is manipulated. This research contributes to the field of community detection by introducing a signature-based similarity measure, offering an alternative to conventional correlation matrices.

The ferroelectric nematic phase (NF) is a recently discovered phase of matter in which the orientational order of the conventional nematic liquid crystal state is augmented with polar order. Atomistic simulations suggest that the polar NF phase would be denser than conventional nematics owing to contributions from polar order. Using an oscillating U-tube densitometer, we obtain detailed temperature-dependent density values for a selection of conventional liquid crystals with excellent agreement with earlier reports. Having demonstrated the validity of our method, we then record density as a function of temperature for M5, a novel room-temperature ferroelectric nematic material. We present the first experimental density data for a NF material as well as density data for a nematic that has not previously been reported. We find that the room-temperature NF material shows a large (>1.3 g cm3) density at all temperatures studied, with an increase in density at phase transitions. The magnitude of the increase for the intermediate splay-ferroelectric nematic (NX-NF) transition is an order of magnitude smaller than the isotropic-nematic (I-N) transition. Present results may be typical of ferroelectric nematic materials, potentially guiding material development, and is especially relevant for informing ongoing studies into this emerging class of materials.

The Consumer Financial Protection Bureau defines the notion of payoff amount as the amount that has to be payed at a particular time in order to completely pay off the debt, in case the lender intends to pay off the loan early, way before the last installment is due (CFPB 2020). This amount is well-understood for loans at compound interest, but much less so when simple interest is used. Recently, Aretusi and Mari (2018) have proposed a formula for the payoff amount for loans at simple interest. We assume that the payoff amounts are established contractually at time zero, whence the requirement that no arbitrage may arise this way The first goal of this paper is to study this new formula and derive it within a model of a loan market in which loans are bought and sold at simple interest, interest rates change over time, and no arbitrage opportunities exist. The second goal is to show that this formula exhibits a behaviour rather different from the one which occurs when compound interest is used. Indeed, we show that, if the installments are constant and if the interest rate is greater than a critical value (which depends on the number of installments), then the sequence of the payoff amounts is increasing before a certain critical time, and will start decreasing only thereafter. We also show that the critical value is decreasing as a function of the number of installments. For two installments, the critical value is equal to the golden section. The third goal is to introduce a more efficient polynomial notation, which encodes a basic tenet of the subject: Each amount of money is embedded in a time position (to wit: The time when it is due). The model of a loan market we propose is naturally linked to this new notation.

We calculate the soft function using lattice QCD in the framework of large momentum effective theory incorporating the one-loop perturbative contributions. The soft function is a crucial ingredient in the lattice determination of light cone objects using transverse-momentum-dependent (TMD) factorization. It consists of a rapidity-independent part called intrinsic soft function and a rapidity-dependent part called Collins-Soper kernel. We have adopted appropriate normalization when constructing the pseudo-scalar meson form factor that is needed in the determination of the intrinsic part and applied Fierz rearrangement to suppress the higher-twist effects. In the calculation of CS kernel we consider a CLS ensemble other than the MILC ensemble used in a previous study. We have also compared the applicability of determining the CS kernel using quasi TMDWFs and quasi TMDPDFs. As an example, the determined soft function is used to obtain the physical TMD wave functions (WFs) of pion and unpolarized iso-vector TMD parton distribution functions (PDFs) of proton., Comment: 24 pages, 19 figures, published version

The resolution and contrast of microscope imaging is often affected by aberrations introduced by imperfect optical systems and inhomogeneous refractive structures in specimens. Adaptive optics (AO) compensates these aberrations and restores diffraction limited performance. A wide range of AO solutions have been introduced, often tailored to a specific microscope type or application. Until now, a universal AO solution -- one that can be readily transferred between microscope modalities -- has not been deployed. We propose versatile and fast aberration correction using a physics-based machine learning (ML) assisted wavefront-sensorless AO control method. Unlike previous ML methods, we used a bespoke neural network (NN) architecture, designed using physical understanding of image formation, that was embedded in the control loop of the microscope. The approach means that not only is the resulting NN orders of magnitude simpler than previous NN methods, but the concept is translatable across microscope modalities. We demonstrated the method on a two-photon, a three-photon and a widefield three-dimensional (3D) structured illumination microscope. Results showed that the method outperformed commonly-used modal-based sensorless AO methods. We also showed that our ML-based method was robust in a range of challenging imaging conditions, such as extended 3D sample structures, specimen motion, low signal to noise ratio and activity-induced fluorescence fluctuations. Moreover, as the bespoke architecture encapsulated physical understanding of the imaging process, the internal NN configuration was no-longer a ``black box'', but provided physical insights on internal workings, which could influence future designs., Comment: 27 pages, 12 figures

We introduce a novel computational unit for neural networks that features multiple biases, challenging the traditional perceptron structure. This unit emphasizes the importance of preserving uncorrupted information as it is passed from one unit to the next, applying activation functions later in the process with specialized biases for each unit. Through both empirical and theoretical analyses, we show that by focusing on increasing biases rather than weights, there is potential for significant enhancement in a neural network model's performance. This approach offers an alternative perspective on optimizing information flow within neural networks. See source code at https://github.com/CuriosAI/dac-dev., Comment: Major rewriting. Supersedes v1 and v2. Focusing on the fact that not all parameters are born equal: biases can be more important than weights. Accordingly, new title and new abstract, and many more experiments on fully connected architectures. This is the extended version of the paper published at WACV 2024

Nanoribbons - nanometer wide strips of a two-dimensional material - are a unique system in condensed matter physics. They combine the exotic electronic structures of low-dimensional materials with an enhanced number of exposed edges, where phenomena including ultralong spin coherence times, quantum confinement and topologically protected states can emerge. An exciting prospect for this new material concept is the potential for both a tunable semiconducting electronic structure and magnetism along the nanoribbon edge. This combination of magnetism and semiconducting properties is the first step in unlocking spin-based electronics such as non-volatile transistors, a route to low-energy computing, and has thus far typically only been observed in doped semiconductor systems and/or at low temperatures. Here, we report the magnetic and semiconducting properties of phosphorene nanoribbons (PNRs). Static (SQUID) and dynamic (EPR) magnetization probes demonstrate that at room temperature, films of PNRs exhibit macroscopic magnetic properties, arising from their edge, with internal fields of ~ 250 to 800 mT. In solution, a giant magnetic anisotropy enables the alignment of PNRs at modest sub-1T fields. By leveraging this alignment effect, we discover that upon photoexcitation, energy is rapidly funneled to a dark-exciton state that is localized to the magnetic edge and coupled to a symmetry-forbidden edge phonon mode. Our results establish PNRs as a unique candidate system for studying the interplay of magnetism and semiconducting ground states at room temperature and provide a stepping-stone towards using low-dimensional nanomaterials in quantum electronics., Comment: 18 pages, 4 figures

We present a first lattice QCD calculation of the unpolarized nucleon's isovector transverse-momentum-dependent parton distribution functions (TMDPDFs), which are essential to predict observables of multi-scale, semi-inclusive processes in the standard model. We use a $N_f=2+1+1$ MILC ensemble with valence clover fermions on a highly improved staggered quark (HISQ) sea to compute the quark momentum distributions in a large-momentum nucleon on the lattice. The state-of-the-art techniques in renormalization and extrapolation in the correlation distance on the lattice are adopted. {The perturbative kernel up to next-to-next-to-leading order is taken into account}, and the dependence on the pion mass and the hadron momentum is explored. Our results are qualitatively comparable with phenomenological TMDPDFs, which provide an opportunity to predict high energy scatterings from first principles., Comment: 16 pages, 20 figures

This paper aims to investigate the possibility of using artificial intelligence (AI) to resolve the legal issues raised by the Covid-19 emergency about the fate of continuing execution contracts, or those with deferred or periodic execution, as well as, more generally, to deal with exceptional events and contingencies. We first study whether the Italian legal system allows for ''maintenance'' remedies to cope with contingencies and to avoid the termination of the contract, while ensuring effective protection of the interests of both parties. We then give a complete and technical description of an AI-based predictive framework, aimed at assisting both the Magistrate (in the course of litigation) and the parties themselves (in out-of-court proceedings) in the redetermination of the rent of commercial lease contracts. This framework, called A.I.A.Co. for Artificial Intelligence for contract law Against Covid-19, has been developed under the Italian grant ''Fondo Integrativo Speciale per la Ricerca''., Comment: Presented at CRCL 2022: Computational 'law' on edge, https://www.cohubicol.com/about/conference-crcl-2022. Version v2 is a major revision after the conference feedback

Understanding (de)lithiation heterogeneities in battery materials is key to ensuring optimal electrochemical performance and developing better energy storage devices. However, this remains challenging due to the complex three dimensional morphology of microscopic electrode particles, the involvement of both solid and liquid phase reactants, and range of relevant timescales (seconds to hours). Here, we overcome this problem and demonstrate the use of bench-top laser scanning confocal microscopy for simultaneous three-dimensional operando measurement of lithium ion dynamics in single particles, and the electrolyte, in batteries. We examine two technologically important cathode materials that are known to suffer from intercalation heterogeneities: LixCoO2 and LixNi0.8Mn0.1Co0.1O2. The single-particle surface-to-core transport velocity of Li-phase fronts, and volume changes - as well as their inter-particle heterogeneity - are captured as a function of C-rate, and benchmarked to previous ensemble measurements. Additionally, we visualise heterogeneities in the bulk and at the surface of particles during cycling, and image the formation of spatially non-uniform concentration gradients within the liquid electrolyte. Importantly, the conditions under which optical imaging can be performed inside absorbing and multiply scattering materials such as battery intercalation compounds are outlined., Comment: 29 pages, 6 figures

In the last years, the DeepMind algorithm AlphaZero has become the state of the art to efficiently tackle perfect information two-player zero-sum games with a win/lose outcome. However, when the win/lose outcome is decided by a final score difference, AlphaZero may play score-suboptimal moves because all winning final positions are equivalent from the win/lose outcome perspective. This can be an issue, for instance when used for teaching, or when trying to understand whether there is a better move. Moreover, there is the theoretical quest for the perfect game. A naive approach would be training an AlphaZero-like agent to predict score differences instead of win/lose outcomes. Since the game of Go is deterministic, this should as well produce an outcome-optimal play. However, it is a folklore belief that "this does not work". In this paper, we first provide empirical evidence for this belief. We then give a theoretical interpretation of this suboptimality in general perfect information two-player zero-sum game where the complexity of a game like Go is replaced by the randomness of the environment. We show that an outcome-optimal policy has a different preference for uncertainty when it is winning or losing. In particular, when in a losing state, an outcome-optimal agent chooses actions leading to a higher score variance. We then posit that when approximation is involved, a deterministic game behaves like a nondeterministic game, where the score variance is modeled by how uncertain the position is. We validate this hypothesis in AlphaZero-like software with a human expert., Comment: Published at 2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA). This version (v2) is a major revision and superseeds version v1

In 2018, California Governor Gavin Newsom set an ambitious goal of reaching 500,000 active apprentices by 2029. He did so because apprenticeship is a proven model and clear strategy for connecting Californians to good jobs. It has a strong track record for workers and employers in the skilled trades and in firefighting, but apprenticeship innovations are also emerging in new sectors, such as health care, information technology, advanced manufacturing, and education, opening up access to good jobs for workers and students through a combination of paid, on-the-job training, and debt-free postsecondary credentials. Access to a good job and family sustaining career is a critical cornerstone for building a more inclusive economy for communities that have always been cut off, or who today find themselves today increasingly distant from opportunities to share in the state's prosperity. But paths to a good job for California's workers and students have only narrowed. The ongoing impact of COVID-19 is making the important work of creating more clear pathways to good jobs and care simultaneously more complicated, more urgent, and more expansive in scale. COVID-19's impacts on the state's economy, especially on the small businesses which employ nearly half of Californian workers, mean more uncertainty for workers in the coming years, especially for those who were already on the fringes of a changing labor market. Like all other states, California will have many policy priorities to manage with diminished budget resources. Reaching a half a million apprentices by 2029 will rest on a strong base of apprenticeship in the building trades and firefighting, where the majority of the state's apprentices are concentrated today. But it will also demand a more expansive and inclusive apprenticeship system, where today nearly 70 percent of Californian apprentices are racial minorities, but only 7 percent are women. Apprenticeship is a strategy that can and should be applied much more broadly across California's workforce. Despite its ambition, the 10-year goal of a 500,000-apprentice system creates the necessary time and space for policymakers to evaluate and implement the necessary actions. At the same time, many of the policy ideas outlined in this report are mutually reinforcing, and as such will have the greatest potential when taken together as part of a comprehensive roadmap for expanding the availability of high-quality apprenticeship opportunities in new and emerging nontraditional industry sectors.