Applied Spectral Imaging (United States)

Deep learning (DL)/artificial intelligence (AI) workloads are known to be computationally demanding, and recent advances with technologies like large language models (LLMs) require scaling to large pools of GPUs to hit performance targets. Often, large HPC/AI workloads will rely on Ethernet or InfiniBand (IB) networking to scale across multi-node clusters once the resources within a node are insufficient. The challenge is that using multi-node resources is difficult. Workloads need to adopt technologies like MPI to communicate across nodes, and there are various challenges associated with managing shared file systems and software across a cluster of nodes. Advances in composable disaggregated infrastructure (CDI) provide an alternative. They allow system designers to scale the system bus to entire racks of resources, allowing multi-node scaling within the convenience of a single node. However, there are technical differences between scaling system resources and scaling network resources, and the constraints imposed by technologies like PCIe provide an interesting set of trade-offs that need to be considered when designing systems. In this work, we present a system bus optimized for AI workloads. Our proposed design leverages a rail-optimized topology to achieve strong performance for LLM applications. We demonstrate how the same topology and software routing optimization deployed on Ethernet/IB networks can be applied to large-scale system bus deployments. Our proposed design can achieve 3.7x better collective performance compared to a more traditional system bus design, and application-level evaluations demonstrate how intelligent topology design can provide a speedup of up to 1.8x in LLM inference and a performance improvement of up to 12% in LLM training throughput.

Abstract We investigate induced seismicity associated with a hydraulic stimulation campaign performed in 2020 in the 5.8 km deep geothermal OTN‐2 well near Helsinki, Finland as part of the St1 Deep Heat project. A total of 2,875 m3 of fresh water was injected during 16 days at well‐head pressures <70 MPa and with flow rates between 400 and 1,000 L/min. The seismicity was monitored using a high‐resolution seismic network composed of 10 borehole geophones surrounding the project site and a borehole array of 10 geophones located in adjacent OTN‐3 well. A total of 6,121 induced earthquakes with local magnitudes were recorded during and after the stimulation campaign. The analyzed statistical parameters include magnitude‐frequency b‐value, interevent time and interevent time ratio, as well as magnitude correlations. We find that the b‐value remained stationary for the entire injection period suggesting limited stress build‐up or limited fracture network coalescence in the reservoir. The seismicity during the stimulation neither shows signatures of magnitude correlations, nor temporal clustering or anticlustering beyond those arising from varying injection rates. The interevent time statistics are characterized by a Poissonian time‐varying distribution. The calculated parameters indicate no earthquake interaction. Focal mechanisms suggest that the injection activated a spatially distributed network of similarly oriented fractures. The seismicity displays stable behavior with no signatures pointing toward a runaway event. The cumulative seismic moment is proportional to the cumulative hydraulic energy and the maximum magnitude is controlled by injection rate. The performed study provides a base for implementation of time‐dependent probabilistic seismic hazard assessment for the project site.

We outline the design and manufacturing approach employed to construct microwave lenses with arbitrary outlines. The lens is fabricated using multi-jet fusion which is suitable for mass production and allows inhomogeneous variation in dielectric constants. The technique is illustrated by making a microwave lens that provides the same performance as a reflector antenna but has the appearance of a birdhouse.

The incidence of cytogenetic abnormalities in childhood de novo acute myeloid leukaemia (AML) and its prognostic significance was assessed in an Israeli paediatric referral centre. Cytogenetic analysis was successful in 86 of 97 children (< 20 years of age) diagnosed between 1988 and 2002 with de novo AML. Fluorescence in situ hybridization analysis detected new information in 11 of them, leading to reassignment in cytogenetic group classification. The incidence of the various cytogenetic subgroups was as follows: normal - 9%; t(11q23) - 22%; t(8;21) - 13%; t(15;17) - 8%; inv(16) - 3.4%; abn(3q) - 4.6%; 7/7q-(sole or main) - 5.8%; del(9q)(sole) and +21(sole) - 4.6% each; t(8;16) - 2.3%; t(6;9), t(1;22), +8(sole) - 1.1% each; and miscellaneous - 18%. The overall survival (OS) and event-free survival (EFS) (4 years) for 94 patients treated with the modified Berlin-Frankfürt-Münster (BFM) AML protocols (non-irradiated) were 59.9% (SE = 5%) and 55.7% (SE = 5%), respectively, and for the favourable t(8;21), t(15;17) and inv(16), OS was 60% (SE = 15%), 83% (SE = 15%) and 100% respectively. For the normal group it was 62% (SE = 17%), miscellaneous 64% (SE = 12%), t(11q23) 44.6% (SE = 11%) and of the -7/7q-, del(9q)(sole) or t(6;9), none had survived at 4 years. The incidence of cytogenetic subgroups in the Israeli childhood AML population and their outcome were similar to other recently reported paediatric series. Cytogenetic abnormalities still carry clinical relevance for treatment stratification in the context of modern chemotherapy.

Recently, spectral imaging has been successfully applied to 24-color fluorescence in situ hybridization (FISH), and it has evolved into a new powerful method for the analysis of structural and numerical chromosomal aberrations on metaphase spreads (spectral karyotyping) For cytological specimens it is often impossible to use fluorescence microscopy because of (fixation-induced) auto-fluorescence or prior cytological staining. Also, re-examination of archived fluorescent specimens is hampered by fading. However, the need for multi-parameter cytochemical analysis remains when rare or unique material is to be studied in research or clinical diagnosis. Multi-color bright-field microscopy using enzyme precipitates has become feasible for in situ hybridization as well as immunohistochemical detection. Conventional transmission microscopy optics have allowed up to three targets to be detected simultaneously. In parallel to 24-color FISH, we explored the potential of spectral imaging to increase the number of parameters to be analyzed in bright-field microscopy.