Abstract: Aspects described herein may allow for the application of stochastic gradient boosting techniques to the training of deep neural networks by disallowing gradient back propagation from examples that are correctly classified by the neural network model while still keeping correctly classified examples in the gradient averaging. Removing the gradient contribution from correctly classified examples may regularize the deep neural network and prevent the model from overfitting. Further aspects described herein may provide for scheduled boosting during the training of the deep neural network model conditioned on a mini-batch accuracy and/or a number of training iterations. The model training process may start un-boosted, using maximum likelihood objectives or another first loss function.

Abstract: A pipe liner installation recovery apparatus is configured for completing installation of a pipe liner such as an ultraviolet light cure-in-place-pipe liner installed within a host pipe. The pipe liner repair apparatus comprises a film feeder tank, a roller disposed within the tank, an outlet, and an inlet port. A roll of replacement film for repairing the pipe liner is installed on the roller. A fluid supply is connected to the inlet port, such that fluid from the fluid supply is directed into the film feeder tank. Under the pressure of the fluid directed into the film feeder tank, the replacement film contained on the roller, is driven through the film feeder tank, into the outlet, into the mouth of the pipe liner at a first end portion of the pipe liner, and through the pipe liner to a second end portion of the pipe liner.

Abstract: The invention provides, e.g., a recombinant virus comprising (i) a modified TATA box-based promoter, and/or (ii) a modified CAAT box-based promoter operably linked to a gene, wherein the modified TATA box-based promoter and/or modified CAAT box-based promoter lacks a functional TATA box and/or CAAT box and permit selective expression of the gene in a hyperproliferative cell. The recombinant viruses can be used to treat cell proliferative diseases and disorders, including certain forms of cancer.

Abstract: Aspects discussed herein may relate to methods and techniques for embedding constrained and unconstrained optimization programs as layers in a neural network architecture. Systems are provided that implement a method of solving a particular optimization problem by a neural network architecture. Prior systems required use of external software to pre-solve optimization programs so that previously determined parameters could be used as fixed input in the neural network architecture. Aspects described herein may transform the structure of common optimization problems/programs into forms suitable for use in a neural network. This transformation may be invertible, allowing the system to learn the solution to the optimization program using gradient descent techniques via backpropagation of errors through the neural network architecture. Thus these optimization layers may be solved via operation of the neural network itself.

Abstract: The invention provides for a recombinant adenoviral vector comprising a recombinant oncolytic adenovirus which has: (1) a modified transcription regulatory sequence wherein the adenoviral vector is transcriptionally active in cancer cells and/or hyperproliferative cells and transcriptionally attenuated in normal cells, and (2) a transgene encoding one or more cancer antigens that are specific to a subject. The invention further provides a method of producing said recombinant adenoviral vector. The recombinant oncolytic adenoviral vector described above is used in methods of stimulating a heightened immune response against a cancer antigen in a subject or a method of treating cancer in a subject.

Abstract: The present disclosure relates to compositions and methods for producing a recombinant virus, e.g., a recombinant oncolytic adenovirus, using an A549 host cell.

Abstract: Disclosed herein are recombinant adenoviruses with one or more nucleotide sequences inserted between two viral transcription units, formulations comprising the recombinant adenoviruses, and methods of treatment using the recombinant adenoviruses. In some embodiments, the one or more nucleotide sequences are inserted in an IX-E2 insertion site and/or an L5-E4 insertion site.

Abstract: Provided is a fusion protein, e.g., a cytokine receptor fusion protein, e.g., a TGF? trap, with a novel linker sequence to permit the fusion protein to functionally optimally, e.g., to permit a cytokine receptor portion of a cytokine receptor fusion protein to bind optimally to its target cytokine. The fusion proteins, or expression vectors encoding for the fusion proteins, e.g., oncolytic adenoviral expression vectors, can be used to treat cell proliferative diseases and disorders, including certain forms of cancer and inflammatory disorders.

Abstract: Aspects discussed herein may relate to methods and techniques for embedding constrained and unconstrained optimization programs as layers in a neural network architecture. Systems are provided that implement a method of solving a particular optimization problem by a neural network architecture. Prior systems required use of external software to pre-solve optimization programs so that previously determined parameters could be used as fixed input in the neural network architecture. Aspects described herein may transform the structure of common optimization problems/programs into forms suitable for use in a neural network. This transformation may be invertible, allowing the system to learn the solution to the optimization program using gradient descent techniques via backpropagation of errors through the neural network architecture. Thus these optimization layers may be solved via operation of the neural network itself.

Abstract: The present disclosure relates to compositions and methods for producing a recombinant virus, e.g., a recombinant oncolytic adenovirus, using an A549 host cell.

Abstract: Provided is a fusion protein, e.g., a cytokine receptor fusion protein, e.g., a TGF? trap, with a novel linker sequence to permit the fusion protein to functionally optimally, e.g., to permit a cytokine receptor portion of a cytokine receptor fusion protein to bind optimally to its target cytokine. The fusion proteins, or expression vectors encoding for the fusion proteins, e.g., oncolytic adenoviral expression vectors, can be used to treat cell proliferative diseases and disorders, including certain forms of cancer and inflammatory disorders.

Abstract: The invention relates to a recombinant adenovirus comprising two or more therapeutic transgenes, e.g., two components of a heterodimeric cytokine, separated by a cleavable linker.

Abstract: Aspects described herein may allow for the application of stochastic gradient boosting techniques to the training of deep neural networks by disallowing gradient back propagation from examples that are correctly classified by the neural network model while still keeping correctly classified examples in the gradient averaging. Removing the gradient contribution from correctly classified examples may regularize the deep neural network and prevent the model from overfitting. Further aspects described herein may provide for scheduled boosting during the training of the deep neural network model conditioned on a mini-batch accuracy and/or a number of training iterations. The model training process may start un-boosted, using maximum likelihood objectives or another first loss function.

Abstract: The invention relates to a method for producing a recombinant virus, e.g., a recombinant oncolytic adenovirus, using an A549 host cell.

Abstract: Aspects discussed herein may relate to methods and techniques for embedding constrained and unconstrained optimization programs as layers in a neural network architecture. Systems are provided that implement a method of solving a particular optimization problem by a neural network architecture. Prior systems required use of external software to pre-solve optimization programs so that previously determined parameters could be used as fixed input in the neural network architecture. Aspects described herein may transform the structure of common optimization problems/programs into forms suitable for use in a neural network. This transformation may be invertible, allowing the system to learn the solution to the optimization program using gradient descent techniques via backpropagation of errors through the neural network architecture. Thus these optimization layers may be solved via operation of the neural network itself.

Abstract: Systems, apparatuses, and methods are described for inverting neural embeddings. One or more forward neural embeddings associated with meanings, features, and/or characteristics of data samples may be generated for one or more data samples. One or more inverse neural embeddings associated with the one or more forward neural embeddings may be determined. One or more inverse feature sets for the one or more inverse neural embeddings may be generated.

Abstract: A wearable health management system includes a flexible member configured to be worn on an affected area by a patient. At least one actuator is operably coupled to the flexible member. The at least one actuator is configured to be adjusted between a deployed state and a non-deployed state. At least one of a photoplethysmogram sensor and a bioimpedance sensor is coupled to the flexible member to obtain one or more health metrics from the patient. A controller is in communication with the at least one actuator. The controller is configured to adjust the at least one actuator to the deployed state to provide a selected pressure to the affected area.

Abstract: Aspects described herein may allow for the application of stochastic gradient boosting techniques to the training of deep neural networks by disallowing gradient back propagation from examples that are correctly classified by the neural network model while still keeping correctly classified examples in the gradient averaging. Removing the gradient contribution from correctly classified examples may regularize the deep neural network and prevent the model from overfitting. Further aspects described herein may provide for scheduled boosting during the training of the deep neural network model conditioned on a mini-batch accuracy and/or a number of training iterations. The model training process may start un-boosted, using maximum likelihood objectives or another first loss function.

Abstract: A wearable health management system includes a flexible member configured to be worn on an affected area by a patient. At least one actuator is operably coupled to the flexible member. The at least one actuator is configured to be adjusted between a deployed state and a non-deployed state. At least one of a photoplethysmogram sensor and a bioimpedance sensor is coupled to the flexible member to obtain one or more health metrics from the patient. A controller is in communication with the at least one actuator. The controller is configured to adjust the at least one actuator to the deployed state to provide a selected pressure to the affected area.

Abstract: Provided is a fusion protein, e.g., a cytokine receptor fusion protein, e.g., an IL-10 trap, with a novel linker sequence to permit the fusion protein to functionally optimally, e.g., to permit a cytokine receptor portion of a cytokine receptor fusion protein to bind optimally to its target cytokine. The fusion protein, or an expression vector encoding for the fusion proteins, can be used to treat cell proliferative diseases and disorders, including certain forms of cancer and inflammatory disorders.