Abstract: Methods, systems, and apparatus for predicting future trajectories of agents in an environment. In one aspect, a system comprises one or more computers configured to receive sensor data of an environment having one or more agents. The system decomposes the sensor data into a plurality of scene elements in the environment, and the system generates multiple tokens including a respective token for each respective scene element of the multiple scene elements in the environment. The system processes the multiple tokens using a decoder model to generate a respective predicted trajectory for the one or more agents in the environment.

Abstract: A method, a use, a device, an electronic equipment and a storage medium for measuring a polarization uniformity of non-uniformly totally polarized beams, which comprises the following steps: inputting a Stokes parameter distribution of a cross section of non-uniformly totally polarized beams; calculating the Stokes parameter distribution to correspond to a Poincare sphere; performing surface fitting on a polarization state distribution on the Poincare sphere through a spatial triangle surface fitting algorithm; calculating a sum SD of the areas of the fitted triangular surfaces; dividing SD by a total area S0 of the unit Poincare sphere to obtain get the polarization uniformity ? of the non-uniformly totally polarized beams.

Abstract: An electrically controlled depolarizer based on a crossed-slit waveguide (3) includes a horizontal-slit waveguide (1), a 45-degree polarization rotation waveguide (2), a pair of modulation electrodes (4) and the crossed-slit waveguide (3).

Abstract: The present disclosure provides a driving circuit, a driving IC, and a driving system, relating to the technical field of electronic circuits. The driving circuit comprises a control module and a driving signal output module, the control module is electrically connected to the driving signal output module, and the driving signal output module is configured to be electrically connected to a to-be-driven device, wherein the driving signal output module comprises at least two transistors, and the at least two transistors are epitaxially grown on the same substrate; and the control module is configured to control a closed state of the at least two transistors, so as to control an operation state of the to-be-driven device.

Abstract: A chip-type three-dimensional magnetic field sensor includes a light source (1), an input straight waveguide (2), a polarization beam splitting waveguide (3), a 1:1 power beam splitter (4), three 1:2 type Y waveguides, three 2:1 type Y waveguides, three output straight waveguides, three magneto-optical waveguides and three photodetectors. The light source (1) outputs broad-spectrum depolarized light into the input straight waveguide (2), and then the light is divided into TE (transverse electric) polarized light and TM (transverse magnetic) polarized light. The TE polarized light is divided into two beams of TE polarized branch light. The TM polarized light is divided into two beams of TM polarized branch light. One of the two beams of TM polarized branch light is divided into two beams of first TM polarized sub-branch light. Another of the two beams of TM polarized branch light is divided into two beams of second TM polarized sub-branch light.

Abstract: The present disclosure provides a driving circuit, a driving IC, and a driving system, relating to the technical field of electronic circuits. The driving circuit comprises a control module and a driving signal output module, the control module is electrically connected to the driving signal output module, and the driving signal output module is configured to be electrically connected to a to-be-driven device, wherein the driving signal output module comprises at least two transistors, and the at least two transistors are epitaxially grown on the same substrate; and the control module is configured to control a closed state of the at least two transistors, so as to control an operation state of the to-be-driven device.

Abstract: An electrically controlled depolarizer based on a crossed-slit waveguide (3) includes a horizontal-slit waveguide (1), a 45-degree polarization rotation waveguide (2), a pair of modulation electrodes (4) and the crossed-slit waveguide (3).

Abstract: A chip-type three-dimensional magnetic field sensor includes a light source (1), an input straight waveguide (2), a polarization beam splitting waveguide (3), a 1:1 power beam splitter (4), three 1:2 type Y waveguides, three 2:1 type Y waveguides, three output straight waveguides, three magneto-optical waveguides and three photodetectors. The light source (1) outputs broad-spectrum depolarized light into the input straight waveguide (2), and then the light is divided into TE (transverse electric) polarized light and TM (transverse magnetic) polarized light. The TE polarized light is divided into two beams of TE polarized branch light. The TM polarized light is divided into two beams of TM polarized branch light. One of the two beams of TM polarized branch light is divided into two beams of first TM polarized sub-branch light. Another of the two beams of TM polarized branch light is divided into two beams of second TM polarized sub-branch light.

Abstract: Methods and systems are provided for ranking images against queries. A visual modality ranking of visual features of a digital image against a query is generated. A language modality ranking of text features of text associated with the digital image against the query is also generated. A multi-modal neural network determines importance weightings of the language modality ranking and the visual modality ranking against the query. The visual modality ranking and the language modality ranking are combined into a multi-modal ranking of the digital image against the query based on the importance weightings. The digital image is provided as a search result of the query based on the multi-modal ranking.

Abstract: Provided is an IL-15 heterodimeric protein, such as an IL-15/IL-15R? heterodimeric protein. The IL-15 heterodimeric protein contains protein (I) and protein (II). Protein (I) is recombinantly produced by combining IL-15 or a variant thereof with a first Fc variant, and protein (II) is a second Fc variant, or recombinantly produced by combining IL-15R? or a variant thereof with a second Fc variant. The first Fc variant and the second Fc variant are preferably linked via a “Knob-into-Hole” mode. Also provided are methods of using the IL-15 heterodimeric proteins to modulate an immune response. The provided IL-15 heterodimeric proteins have significant anti-tumor activity, better biological activity and prolonged in vivo half-life compared to the IL-15 molecule alone.

Abstract: Methods and systems are provided for ranking images against queries. A visual modality ranking of visual features of a digital image against a query is generated. A language modality ranking of text features of text associated with the digital image against the query is also generated. A multi-modal neural network determines importance weightings of the language modality ranking and the visual modality ranking against the query. The visual modality ranking and the language modality ranking are combined into a multi-modal ranking of the digital image against the query based on the importance weightings. The digital image is provided as a search result of the query based on the multi-modal ranking.

Abstract: Described herein are systems and methods for generating and using attention-based deep learning architectures for visual question answering task (VQA) to automatically generate answers for image-related (still or video images) questions. To generate the correct answers, it is important for a model's attention to focus on the relevant regions of an image according to the question because different questions may ask about the attributes of different image regions. In embodiments, such question-guided attention is learned with a configurable convolutional neural network (ABC-CNN). Embodiments of the ABC-CNN models determine the attention maps by convolving image feature map with the configurable convolutional kernels determined by the questions semantics. In embodiments, the question-guided attention maps focus on the question-related regions and filters out noise in the unrelated regions.

Abstract: Described herein are systems and methods for generating and using attention-based deep learning architectures for visual question answering task (VQA) to automatically generate answers for image-related (still or video images) questions. To generate the correct answers, it is important for a model's attention to focus on the relevant regions of an image according to the question because different questions may ask about the attributes of different image regions. In embodiments, such question-guided attention is learned with a configurable convolutional neural network (ABC-CNN). Embodiments of the ABC-CNN models determine the attention maps by convolving image feature map with the configurable convolutional kernels determined by the questions semantics. In embodiments, the question-guided attention maps focus on the question-related regions and filters out noise in the unrelated regions.

Abstract: Provided is an IL-15 heterodimeric protein, such as an IL-15/IL-15R? heterodimeric protein. The IL-15 heterodimeric protein contains protein (I) and protein (II). Protein (I) is recombinantly produced by combining IL-15 or a variant thereof with a first Fc variant, and protein (II) is a second Fc variant, or recombinantly produced by combining IL-15R? or a variant thereof with a second Fc variant. The first Fc variant and the second Fc variant are preferably linked via a “Knob-into-Hole” mode. Also provided are methods of using the IL-15 heterodimeric proteins to modulate an immune response. The provided IL-15 heterodimeric proteins have significant anti-tumor activity, better biological activity and prolonged in vivo half-life compared to the IL-15 molecule alone.

Abstract: Methods and systems for detecting defects on a wafer are provided. One method includes identifying one or more characteristics of first raw output generated for a wafer that correspond to one or more geometrical characteristics of patterned features formed on the wafer and assigning individual output in second raw output generated for the wafer to different segments based on the identified one or more characteristics of the first raw output and based on the individual output in the second raw output and individual output in the first raw output that were generated at substantially the same locations on the wafer such that the one or more geometrical characteristics of the patterned features that correspond to each of the different segments in the second raw output are different.

Abstract: An intelligent vehicle-highway system utilizes a transmitter installed on the vehicle for communicating with modules having memory elements for storing site-specific information, a receiver for receiving an enabling signal from the transceiver on the vehicle, and a transmitter for making the site-specific information available to the operator of the vehicle. The data storage system is installed in the vicinity of the roadway, such as alongside of the roadway, overhead of the roadway, or preferably embedded in the roadway itself so as to ensure communciation with vehicles within predetermined lanes. In a highly advantageous embodiment, the site-specific data can be changed by reprogramming an erasable memory with a memory programmer, which may be installed on a programming vehicle which is authorized to travel on the roadway for the purpose of reprogramming the data within the data storage systems. Such reprogramming may use access codes which prevent tampering with the stored data.