Industrial recommendation systems are typically composed of multiple stages, including retrieval, ranking, and blending. The retrieval stage plays a critical role in generating a high-recall set of candidate items that covers a wide range of diverse user interests. Effectively covering the diverse and long-tail user interests within this stage poses a significant challenge: traditional two-tower models struggle in this regard due to limited user-item feature interaction and often bias towards top use cases. To address these issues, we propose a novel multi-embedding retrieval framework designed to enhance user interest representation by generating multiple user embeddings conditioned on both implicit and explicit user interests. Implicit interests are captured from user history through a Differentiable Clustering Module (DCM), whereas explicit interests, such as topics that the user has followed, are modeled via Conditional Retrieval (CR). These methodologies represent a form of conditioned user representation learning that involves condition representation construction and associating the target item with the relevant conditions. Synergizing implicit and explicit user interests serves as a complementary approach to achieve more effective and comprehensive candidate retrieval as they benefit on different user segments and extract conditions from different but supplementary sources. Extensive experiments and A/B testing reveal significant improvements in user engagements and feed diversity metrics. Our proposed framework has been successfully deployed on Pinterest home feed.

Thermal plasma spraying is an important manufacturing technique that creates a thermal barrier coating to protect the surface underneath from wear, erosion, oxidation and corrosion. In this paper, we develop a new microstructure classification and quantification (MCQ) module that could fully automatically classify and quantify two types of microstructures, globular and interlamellar, in the top coat layer of thermal barrier coating images captured from light microscope. The MCQ module utilizes the masks obtained from the Thermal Barrier Coating Porosity Measurement (TBCPM) framework and applies the morphological filtering of erosion to the images to differentiate globular and interlamellar that are physically distinct and are eroded away at a different rate. The microstructures exhibit complicated spatial relations, for example, a segment might contain microstructures of both types. To deal with this difficulty, we build a forest of hierarchical trees of segments when eroding the images and then traverse this forest of the trees in post order following a carefully designed rule to determine if two neighboring segments should be merged. Our experimental results on the synthesized images with non-overlapping and overlapping segments show an average accuracy of 98.77% and 93.12% respectively.

We introduce a technique to compute probably approximately correct (PAC) bounds on precision and recall for matching algorithms. The bounds require some verified matches, but those matches may be used to develop the algorithms. The bounds can be applied to network reconciliation or entity resolution algorithms, which identify nodes in different networks or values in a data set that correspond to the same entity. For network reconciliation, the bounds do not require knowledge of the network generation process.

Thermal barrier coating, a widely used advanced manufacturing technique, provides thermal insulation and protection of a part surface from inclement environment. In the coating process, microstructures, such as globular and interlamellar, are naturally formed in the coating layer. The amount and types of microstructures greatly influences the property and quality of the coating. This paper introduces a fully automated and innovative approach for classification of microstructures in the thermal barrier coating, which utilizes morphological image analysis and supervised machine learning techniques to differentiate between globular and interlamellar microstructures. The experimental results demonstrate the robustness and effectiveness of the proposed method.

Excessive peak power supply noise (PPSN) causes yield loss problem during test. To reduce PPSN, we proposed a new technique called Capture and Shift Toggle Reduction (CASTR). CASTR performs power reduction during dynamic test compaction so the test length overhead is very small. It also includes pseudo Boolean optimization (PBO) and random-based techniques to improve the results. Experimental results show that we can reduce flip-flop toggles, which is highly correlated with PPSN, by 33.4% during shift mode and 41.2% in capture operation simultaneously for the large ISCAS89 benchmarks.

This paper presents slow-speed testing results for two test chip experiments at Stanford CRC: Murphy and ELF35. The test speed for slow-speed testing is 1/3 of the characterized speed, at which the circuit-under-test (CUT) can operate. At nominal supply voltage, 3 out the 116 defective Murphy CUTs escaped slow-speed testing. In the ELF35 experiment, I out of the 218 defective combinational CUTs escaped slow-speed testing. The experimental data also show that the number of test escapes depends more on the quality of the applied test sets, than on the test speed at which the test sets were applied We also evaluated the effectiveness of VLV testing at characterized speed. Our results show that it missed only one defective Murphy CUT. It detected all the defective ELF35 combinational CUTs. The one defective Murphy CUT that escaped VLV testing at characterized speed is a suspected resistive open. Schmoo characterization results show it can be detected by VLV testing at characterized speed if lot-to-lot re-characterization is done to determine the VLV testing speed.

As the threat of terrorism becomes more prevalent, the need for technological innovations for tighter security rises. Some of the areas where security needs to be increased are at border customs stations, airports, federal buildings, embassies, sporting events, and the like where checkpoints are being set up to inspect vehicles coming onto the premises. The proposed vehicle inspection system uses an at-ray of cameras and mosaicing techniques to generate different perspective views that provide a 3D representation of a vehicle's undercarriage. These mosaics allow a complete inspection of the undercarriage using a graphical user interface without requiring the use of inspection mirrors or the vehicle to drive over a manned inspection bay. The system is designed for simplicity so that operators will not require long training periods.