Abstract: A computer-implemented method for generating video representations utilizing a hierarchical video encoder includes obtaining a video, wherein the video includes a plurality of frames, processing each of the plurality of frames with a machine-learned frame-level encoder model to respectively generate a plurality of frame representations for the plurality of frames, the plurality of frame representations respective to the plurality of frames determining a plurality of segment representations representative of a plurality of video segments including one or more of the plurality of frames, the plurality of segment representations based at least in part on the plurality of frame representations, processing the plurality of segment representations with a machine-learned segment-level encoder model to generate a plurality of contextualized segment representations, determining a video representation based at least in part on the plurality of contextualized segment representations, and providing the video representati

Abstract: A computer-implemented method for generating video representations utilizing a hierarchical video encoder includes obtaining a video, wherein the video includes a plurality of frames, processing each of the plurality of frames with a machine-learned frame-level encoder model to respectively generate a plurality of frame representations for the plurality of frames, the plurality of frame representations respective to the plurality of frames determining a plurality of segment representations representative of a plurality of video segments including one or more of the plurality of frames, the plurality of segment representations based at least in part on the plurality of frame representations, processing the plurality of segment representations with a machine-learned segment-level encoder model to generate a plurality of contextualized segment representations, determining a video representation based at least in part on the plurality of contextualized segment representations, and providing the video representati

Abstract: This invention discloses a model test device and method for curtain grouting and excavation of tunnels in high-temperature, water-rich and weak strata, comprising weak strata simulation system, working condition simulation system, curtain grouting system, and monitoring and data acquisition system. The weak strata simulation system is the main box used to fill the soft soil. The working condition simulation system includes a high geothermal simulation system, a water pressure simulation system with stepwise loading of water pressure and volume, and a three-dimensional confining pressure simulation system with stepwise controllable loading of confining pressure. The invention can realize the whole test process simulation of grouting reinforcement and excavation of tunnels in weak strata effectively under the action of confining pressure and geothermal heat, and it has a positive reference effect on actual soft stratum tunnel curtain grouting and safe excavation.

Abstract: An equipment and application method for sinkage detection and active escape of deep-sea mining vehicle, comprising: the sinkage detection and warning devices and the self-escape equipment. The sinkage detection and warning devices are installed on the left and right sides of each track of the deep-sea mining vehicle, and it monitors the contact conditions between the tracks and the sediment, and the distance from the bottom of the track to the device is measured. The self-escape equipment comprises the active interactive grouting system, which connects to intelligent steering and retractable grouting heads. The intelligent steering and retractable grouting heads are also located on the left and right sides of each track. This invention can detect the complex driving environment of the seafloor automatically and combine the existing grouting technology and materials to achieve active grouting escape in case of excessive sinkage.

Abstract: This invention discloses a model test apparatus and method for shield post-wall grouting and tube sheet uplift under high water pressure conditions, comprising main body box, tension apparatus, shield system, grouting system, ground stress loading system, water injection system, and monitoring system. The main body box comprises the shield shell system and model soil inside. The ground stress loading system and water injection system realize the simulation of the complex underground soil and muddy water environment by applying stress and water pressure to the model soil inside the main box. While the tension apparatus makes the shield shell moving forward, the grouting system injects slurry into the gap between the shield shell and the tube sheet. The integrated model test of visual shield grouting and tube sheet uplift, the real-time multi-physical field responses, and the longitudinal deformation mechanism have been realized and revealed under high water pressure conditions.

Abstract: This invention discloses a double-liquid grouting slurry, its technology and application for super large diameter underwater shield engineering under high water pressure condition. The materials of slurry I are: 35-45 parts of cement clinker; 15-25 parts of slag; 24-35 parts of fly ash; 15-25 parts of steel slag; 5-15 parts of bentonite; 4-10 parts of limestone tailing; 0.3-2.0 parts of water reducing agent; 0.5-2.5 parts of cellulose. The materials of slurry II are: 0.2-3.8 parts of short-cut fiber; 96-99 parts of sodium silicate solution; 0.8-4.8 parts of viscous polymers. This invention generates the double-liquid slurry preparation process including crushing-screening-milling-group mixing-grouped mixing at different speeds, the volume ratio of slurry I and II is 1:1-10:1 during grouting, and the slurry is injected into the shield void through the six-point position technology at the shield tail and 3+2+1 segment splicing synchronous grouting techniques.

Abstract: A synchronous single-liquid grouting slurry, its technology and application for large diameter shield engineering under water-rich, high-pressure and weak soil strata conditions, comprising raw materials: 1050-1200 parts of gold tailing, 420-480 parts of silicate cement clinker, 220-240 parts of fly ash, 45-120 parts of waste clay brick, 65-95 parts of slag, 25-45 parts of limestone tailing, 70-80 parts of steel slag, 30-45 parts of silica fume, 15-22 parts of desulfurized gypsum, and 9-15 parts of quick-setting and early-strength composite additive. The invention controls the d50, d85 and d95 of the material particles as 35-40, 42-48 and 50-55 ?m, respectively. Gold tailing with the particle size of 120-600 ?m being used as the fine aggregate, their volume fractions are 40-60%. The slurry production technique, comprising crushing-sieving-superfine ball milling-homogenization-particle size classification-variable speed mixing being developed.

Abstract: A computer-implemented method for generating video representations utilizing a hierarchical video encoder includes obtaining a video, wherein the video includes a plurality of frames, processing each of the plurality of frames with a machine-learned frame-level encoder model to respectively generate a plurality of frame representations for the plurality of frames, the plurality of frame representations respective to the plurality of frames determining a plurality of segment representations representative of a plurality of video segments including one or more of the plurality of frames, the plurality of segment representations based at least in part on the plurality of frame representations, processing the plurality of segment representations with a machine-learned segment-level encoder model to generate a plurality of contextualized segment representations, determining a video representation based at least in part on the plurality of contextualized segment representations, and providing the video representati

Abstract: A computer-implemented method for generating video representations utilizing a hierarchical video encoder includes obtaining a video, wherein the video includes a plurality of frames, processing each of the plurality of frames with a machine-learned frame-level encoder model to respectively generate a plurality of frame representations for the plurality of frames, the plurality of frame representations respective to the plurality of frames determining a plurality of segment representations representative of a plurality of video segments including one or more of the plurality of frames, the plurality of segment representations based at least in part on the plurality of frame representations, processing the plurality of segment representations with a machine-learned segment-level encoder model to generate a plurality of contextualized segment representations, determining a video representation based at least in part on the plurality of contextualized segment representations, and providing the video representati

Abstract: A proactive networking system and method is disclosed. The network anticipates the user demands in advance and utilizes this predictive ability to reduce the peak to average ratio of the wireless traffic and yield significant savings in the required resources to guarantee certain Quality of Service (QoS) metrics. The system and method focuses on the existing cellular architecture and involves the design and analysis of learning algorithms, predictive resource allocation strategies, and incentive techniques to maximize the efficiency of proactive cellular networks. The system and method further involve proactive peer-to-peer (P2P) overlaying, which leverages the spatial and social structure of the network. Machine learning techniques are applied to find the optimal tradeoff between predictions that result in content being retrieved that the user ultimately never requests, and requests that are not anticipated in a timely manner.

Abstract: A computer-implemented method for generating video representations utilizing a hierarchical video encoder includes obtaining a video, wherein the video includes a plurality of frames, processing each of the plurality of frames with a machine-learned frame-level encoder model to respectively generate a plurality of frame representations for the plurality of frames, the plurality of frame representations respective to the plurality of frames determining a plurality of segment representations representative of a plurality of video segments including one or more of the plurality of frames, the plurality of segment representations based at least in part on the plurality of frame representations, processing the plurality of segment representations with a machine-learned segment-level encoder model to generate a plurality of contextualized segment representations, determining a video representation based at least in part on the plurality of contextualized segment representations, and providing the video representati

Abstract: A proactive networking system and method is disclosed. The network anticipates the user demands in advance and utilizes this predictive ability to reduce the peak to average ratio of the wireless traffic and yield significant savings in the required resources to guarantee certain Quality of Service (QoS) metrics. The system and method focuses on the existing cellular architecture and involves the design and analysis of learning algorithms, predictive resource allocation strategies, and incentive techniques to maximize the efficiency of proactive cellular networks. The system and method further involve proactive peer-to-peer (P2P) overlaying, which leverages the spatial and social structure of the network. Machine learning techniques are applied to find the optimal tradeoff between predictions that result in content being retrieved that the user ultimately never requests, and requests that are not anticipated in a timely manner.

Abstract: Histologic classification of pathology specimens through machine learning is a nascent field which offers tremendous potential to improve cancer medicine. Its utility has been limited, in part because of differences in tissue preparation and the relative paucity of well-annotated images. We introduce tissue recognition, an unsupervised learning problem analogous to human face recognition, in which the goal is to identify individual tumors using a learned set of histologic features. This feature set is the “tissue fingerprint.” Because only specimen identities are matched to fingerprints, constructing an algorithm for producing them is a self-learning task that does not need image metadata annotations. Here, we provide an algorithm for self-learning tissue fingerprints, that, in conjunction with color normalization, can match hematoxylin and eosin stained tissues to one of 104 patients with 93% accuracy.

Abstract: In one embodiment, a method includes receiving a user request for a summarization of a particular type of content objects from a client system associated with a first user, determining one or more modalities associated with the user request, selecting a plurality of content objects of the particular type based on a user profile of the first user, wherein the user profile comprises one or more confidence scores associated with one or more subjects associated with the first user, respectively, and wherein the plurality of content objects are selected based on the one or more confidence scores, generating a summary of each content object based on the user profile and the determined modalities, and sending, to the client system in response to the user request, instructions for presenting the summaries of the plurality of content objects, wherein the summaries are presented via one or more of the determined modalities.

Abstract: A proactive networking system and method is disclosed. The network anticipates the user demands in advance and utilizes this predictive ability to reduce the peak to average ratio of the wireless traffic and yield significant savings in the required resources to guarantee certain Quality of Service (QoS) metrics. The system and method focuses on the existing cellular architecture and involves the design and analysis of learning algorithms, predictive resource allocation strategies, and incentive techniques to maximize the efficiency of proactive cellular networks. The system and method further involve proactive peer-to-peer (P2P) overlaying, which leverages the spatial and social structure of the network. Machine learning techniques are applied to find the optimal tradeoff between predictions that result in content being retrieved that the user ultimately never requests, and requests that are not anticipated in a timely manner.

Abstract: A proactive networking system and method is disclosed. The network anticipates the user demands in advance and utilizes this predictive ability to reduce the peak to average ratio of the wireless traffic and yield significant savings in the required resources to guarantee certain Quality of Service (QoS) metrics. The system and method focuses on the existing cellular architecture and involves the design and analysis of learning algorithms, predictive resource allocation strategies, and incentive techniques to maximize the efficiency of proactive cellular networks. The system and method further involve proactive peer-to-peer (P2P) overlaying, which leverages the spatial and social structure of the network. Machine learning techniques are applied to find the optimal tradeoff between predictions that result in content being retrieved that the user ultimately never requests, and requests that are not anticipated in a timely manner.

Abstract: In one embodiment, a method includes receiving a user request for a summarization of a particular type of content objects from a client system associated with a first user, determining one or more modalities associated with the user request, selecting a plurality of content objects of the particular type based on a user profile of the first user, wherein the user profile comprises one or more confidence scores associated with one or more subjects associated with the first user, respectively, and wherein the plurality of content objects are selected based on the one or more confidence scores, generating a summary of each content object based on the user profile and the determined modalities, and sending, to the client system in response to the user request, instructions for presenting the summaries of the plurality of content objects, wherein the summaries are presented via one or more of the determined modalities.

Abstract: A proactive networking system and method is disclosed. The network anticipates the user demands in advance and utilizes this predictive ability to reduce the peak to average ratio of the wireless traffic and yield significant savings in the required resources to guarantee certain Quality of Service (QoS) metrics. The system and method focuses on the existing cellular architecture and involves the design and analysis of learning algorithms, predictive resource allocation strategies, and incentive techniques to maximize the efficiency of proactive cellular networks. The system and method further involve proactive peer-to-peer (P2P) overlaying, which leverages the spatial and social structure of the network. Machine learning techniques are applied to find the optimal tradeoff between predictions that result in content being retrieved that the user ultimately never requests, and requests that are not anticipated in a timely manner.

Abstract: A proactive networking system and method is disclosed. The network anticipates the user demands in advance and utilizes this predictive ability to reduce the peak to average ratio of the wireless traffic and yield significant savings in the required resources to guarantee certain Quality of Service (QoS) metrics. The system and method focuses on the existing cellular architecture and involves the design and analysis of learning algorithms, predictive resource allocation strategies, and incentive techniques to maximize the efficiency of proactive cellular networks. The system and method further involve proactive peer-to-peer (P2P) overlaying, which leverages the spatial and social structure of the network. Machine learning techniques are applied to find the optimal tradeoff between predictions that result in content being retrieved that the user ultimately never requests, and requests that are not anticipated in a timely manner.

Abstract: An admissions funnel process software application recommends a candidate to a school based on the admission cycle timeline of the school. The admission closing tool receives an admission schedule for a school that includes a set of admission goals for the school such as a desired number of students at each stage of an admission cycle. A candidate profile is compiled for each of a plurality of candidates. For each candidate, the admissions funnel process software application predicts a stage in the admission cycle that the candidate will be, at a future time. For each stage of an admission cycle, the candidates are aggregated by their predicted stages to obtain a predicted admissions cycle for the school. The aggregated number of candidates is compared to the desired number of candidates for that stage. Responsive to the comparing, one or more candidates are selected as leads, recommended and sent to the school.