Search

Recent rapid advancements of machine learning have greatly enhanced the accuracy of prediction models, but most models remain "black boxes", making prediction error diagnosis challenging, especially with outliers. This lack of transparency hinders trust and reliability in industrial applications. Heuristic attribution methods, while helpful, often fail to capture true causal relationships, leading to inaccurate error attributions. Various root-cause analysis methods have been developed using Shapley values, yet they typically require predefined causal graphs, limiting their applicability for prediction errors in machine learning models. To address these limitations, we introduce the Causal-Discovery-based Root-Cause Analysis (CD-RCA) method that estimates causal relationships between the prediction error and the explanatory variables, without needing a pre-defined causal graph. By simulating synthetic error data, CD-RCA can identify variable contributions to outliers in prediction errors by Shapley values. Extensive simulations show CD-RCA outperforms current heuristic attribution methods, and a sensitivity analysis reveals new patterns where Shapley values may misattribute errors, paving the way for more accurate error attribution methods., Comment: 10 pages with 5 figures

In a crystalline solid under mechanical stress, a Frank-Read source is a pinned dislocation segment that repeatedly bows and detaches, generating concentric dislocation loops. We demonstrate that in nematic liquid crystals, an analogous Frank-Read mechanism can generate concentric disclination loops. Using experiment, simulation, and theory, we study a disclination segment pinned between surface defects on one substrate in a nematic cell. Under applied twist of the nematic director, the pinned segment bows and emits a new disclination loop which expands, leaving the original segment intact; loop emission repeats for each additional 180$^\circ$ of applied twist. We present experimental micrographs showing loop expansion and snap-off, numerical simulations of loop emission under both quasistatic and dynamic loading, and theoretical analysis considering both free energy minimization and the balance of competing forces. We find that the critical stress for disclination loop emission scales as the inverse of segment length, and changes as a function of strain rate and temperature, in close analogy to the Frank-Read source mechanism in crystals. Lastly, we discuss how Frank-Read sources could be used to modify microstructural evolution in both passive and active nematics., Comment: 29 pages, 6 figures

Wire-grid polarizer sheets in the terahertz region have been fabricated on flexible substrates by nanoimprint technology. They show ideal polarization property in the terahertz frequency region whereas the cost is very low. Since the wire pitch is far smaller than the wavelength, the effective medium theory agrees well with experimental results. The effective medium theory shows possibility of further improvement of polarization properties by selecting appropriate materials for wire grids.

We show via proof-by-construction that the chain fountain, also known as the Mould effect, is a generic consequence of energy conserving dynamics of linear chains, similar to the extra acceleration observed for hanging or falling chains. Extracting a chain from its unentangled slack sitting on a table is much less energy dissipative than the claim of the classical scenario, postulated by Biggins and Warner [Proc. R. Soc. A 470, 20130689(2014)] as the starting point of their explanation of the chain fountain. As a result, their alleged upward kicks on the chain from the table as the chain is uplifted can at best be one of the possible factors in driving the fountain, if not totally irrelevant. We construct an efficient chain fountain with no upward kicks in Biggins and Warner\'s sense, rather relying on the hanging chain dynamics. Simply put, the centrifugal force at the top of the fountain is responsible for maintaining the fountain. We argue that lateral motion of the chain in the slack plays a decisive role in the Mould effect., Comment: 27 pages, 16 figures ,9 video files

According to the recently discovered 'Law of Urination', mammals, ranging in size from mice to elephants, take, on the average, 21s to urinate. We attempt to gain insights into the physical processes responsible for this uniformity using simple dimensional analysis. We assume that the biological apparatus for urination in mammals simply scales with linear size, and consider the scenarios where the driving force is gravity or elasticity, and where the response is dominated by inertia or viscosity. We ask how the time required for urination depends on the length scale, and find that for the time to be independent of body size, the dominant driving force must be elasticity, and the dominant response viscosity. Our note demonstrates that dimensional analysis can indeed readily give insights into complex physical and biological processes.