Search for: All records

Deep learning-based prediction models for High-Level Synthesis (HLS) of hardware designs often struggle to generalize. In this paper, we study how to close the generalizability gap of these models through pretraining on synthetic data and introduce Iceberg, a synthetic data augmentation approach that expands both large language model (LLM)-generated programs and weak labels of unseen design configurations. Our weak label generation method is integrated with an in-context model architecture, enabling meta-learning from actual and proximate labels. Iceberg improves the geometric mean modeling accuracy by 86.4% when adapt to six real-world applications with few-shot examples and achieves a 2.47× and a 1.12× better offline DSE performance when adapting to two different test datasets. Our open-sourced code is here: https://github.com/UCLA-VAST/iceberg. 

Abstract in attached file 

Abstract We report a new population of outer belt electron acceleration events ranging from hundreds of keV to ∼1.5 MeV that occurred inside the plasmasphere, which we named “Inside Events” (IEs). Based on 6 year observations from Van Allen Probes, we compare the statistical distributions of IEs with electron acceleration events outside the plasmasphere (OEs). We find that most IEs were observed atL < 4.0 at energies below ∼1.5 MeV, with weaker acceleration ratio (<10) and larger event numbers (peaking value reaching >200), compared to stronger but less frequently occurred (peaking event numbers only reaching ∼80) OEs that were mostly observed atL > 4.0. The evolution of electron phase space density of a typical IE shows signature of inward radial diffusion or transport. Our study provides a feasible mechanism for IE, which is the results of the inward radial transport of the electron acceleration in the outer region of outer belt. 

Abstract Previous research has shown that when domain‐general transitional probability (TP) cues to word segmentation are in conflict with language‐specific stress cues, English‐learning 5‐ and 7‐month‐olds rely on TP, whereas 9‐month‐olds rely on stress. In two artificial languages, we evaluated English‐learning infants’ sensitivity to TP cues to word segmentation vis‐a‐vis language‐specific vowel phonotactic (VP) cues—English words do not end in lax vowels. These cues were either consistent or conflicting. When these cues were in conflict, 10‐month‐olds relied on the VP cues, whereas 5‐month‐olds relied on TP. These findings align with statistical bootstrapping accounts, where infants initially use domain‐general distributional information for word segmentation, and subsequently discover language‐specific patterns based on segmented words. Research HighlightsResearch indicates that when transitional probability (TP) conflicts with stress cues for word segmentation, English‐learning 9‐month‐olds rely on stress, whereas younger infants rely on TP.In two artificial languages, we evaluated English‐learning infants’ sensitivity to TP versus vowel phonotactic (VP) cues for word segmentation.When these cues conflicted, 10‐month‐olds relied on VPs, whereas 5‐month‐olds relied on TP.These findings align with statistical bootstrapping accounts, where infants first utilize domain‐general distributional information for word segmentation, and then identify language‐specific patterns from segmented words. 

Abstract in attached file 

Abstract We perform a comprehensive investigation of the statistical distribution of outer belt electron acceleration events over energies from 300 keV to ∼10 MeV regardless of storm activity using 6‐years of observations from Van Allen Probes. We find that the statistical properties of acceleration events are consistent with the characteristic energies of combined local acceleration by chorus waves and inward radial diffusion. While electron acceleration events frequently occur both at <2 MeV atL < 4.0 and at multi‐MeV atL > 4.5, significant acceleration events are confined toL > ∼4.0. By performing superposed epoch analysis of acceleration events during storm and non/weak storm events and comparing their geomagnetic conditions, we reveal the strong correlation (>0.8) between accumulated impacts of substorms as measured by time‐integrated AL (Int(AL)) and the upper flux limit of electron acceleration. While intense storms can provide favorable conditions for efficient acceleration, they are not necessarily required to produce large maximum fluxes. 

Abstract Ultraviolet images of Earth's polar regions obtained by high altitude spacecraft have proved to be immensely useful for documenting numerous features of the aurora and understanding the coupling between Earth's magnetosphere and ionosphere. In this study we have examined images obtained by the far ultraviolet Spectrographic Imager camera on the IMAGE satellite during the first three years of its mission (2000–2002) for comparison with observations of large geomagnetic disturbances (GMDs) by ground‐based magnetometers in eastern Arctic Canada. To our knowledge, this is the first study to investigate the use of high‐altitude imager data to identify the global context of GMDs. We found that rapid auroral motions or localized intensifications visible in these images coincide with regions of largedB/dtas well as localized and closely spaced up/down vertical currents and increased equivalent ionospheric currents, but one of the two events presented did not appear to be related to substorm processes. These magnetic perturbations and currents can appear or disappear in a few tens of seconds, thus highlighting the importance of images with a high cadence. 

Abstract Understanding and forecasting outer radiation belt electron flux dropouts is one of the top concerns in space physics. By constructing Support Vector Machine (SVM) models to predict storm‐time dropouts for both relativistic and ultra‐relativistic electrons overL = 4.0–6.0, we investigate the nonlinear correlations between various driving factors (model inputs) and dropouts (model output) and rank their relative importance. Only time series of geomagnetic indices and solar wind parameters are adopted as model inputs. A comparison of the performance of the SVM models that uses only one driving factor at a time enables us to identify the most informative parameter and its optimal length of time history. Its accuracy and the ability to correctly predict dropouts identifies the SYM‐H index as the governing factor atL = 4.0–4.5, while solar wind parameters dominate the dropouts at higher L‐shells (L = 6.0). Our SVM model also gives good prediction of dropouts during completely out‐of‐sample storms.