Abstract: The technology disclosed provides systems and methods related to preventing exfiltration of training data by feature reconstruction attacks on model instances trained on the training data during a training job. The system comprises a privacy interface that presents a plurality of modulators for a plurality of training parameters. The modulators are configured to respond to selection commands via the privacy interface to trigger procedural calls. The procedural calls modify corresponding training parameters in the plurality of training parameters for respective training cycles in the training job. The system comprises a trainer configured to execute the training cycles in dependence on the modified training parameters. The trainer can determine a performance accuracy of the model instances for each of the executed training cycles.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to a system and method for assigning participants to groups in a clinical trial. The system includes a federated server configured with group assignability data specifying a plurality of groups assignable to participants in a clinical trial and group distribution data specifying distribution of the participants into groups. The groups include at least one placebo group and one or more treatment groups. The system includes an intervention server configured to generate group encryption keys for encrypting the group assignability data. The system includes edge devices of each of the participants. The edge devices are in communication with the federated server.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic biometric identifiers. The method includes generating a scannable code upon receiving a success nonce from a registration server. The registration server can access a user identifier and a hash of at least a signature using the success nonce. The signature can be generated based at least in part upon a biometric identifier of a user. The method includes recreating the hash of the signature stored by the registration server. The method includes generating the scannable code by encrypting the success nonce and the recreated hash. The biometric identifier of the user is generated by feeding a plurality of non-deterministic biometric inputs to a trained machine learning model producing a plurality of feature vectors. The method includes projecting the plurality of feature vectors onto a surface of a unit hyper-sphere and computing a characteristic identity vector representing the user.

Abstract: The technology disclosed relates to a system and method for assigning participants to groups in a clinical trial. The system includes a federated server configured with group assignability data specifying a plurality of groups assignable to participants in a clinical trial and group distribution data specifying distribution of the participants into groups. The groups include at least one placebo group and one or more treatment groups. The system includes an intervention server configured to generate group encryption keys for encrypting the group assignability data. The system includes edge devices of each of the participants. The edge devices are in communication with the federated server.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: A computer-implemented method for automatically identifying subjects at risk of Multiple Sclerosis (MS) includes acquiring a plurality of images of a subject's brain using a Magnetic Resonance Imaging (MRI) scanner. A contrast enhancement process is applied to each image to generate a plurality of contrast-enhanced images. An automated lesion detection algorithm is applied to detect one or more lesions present in the contrast-enhanced images. An automated central vein detection algorithm is applied to detect one or more central veins present in the contrast-enhanced images. An automated paramagnetic rim detection algorithm is applied to detect one or more paramagnetic rims present in the contrast-enhanced images. The patient's risk for MS may then be determined based on the one or more of the lesions, central veins, and paramagnetic rims present in the contrast-enhanced images.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to a system and method for assigning participants to groups in a clinical trial. The system includes a federated server configured with group assignability data specifying a plurality of groups assignable to participants in a clinical trial and group distribution data specifying distribution of the participants into groups. The groups include at least one placebo group and one or more treatment groups. The system includes an intervention server configured to generate group encryption keys for encrypting the group assignability data. The system includes edge devices of each of the participants. The edge devices are in communication with the federated server.

Abstract: The technology disclosed provides systems and methods related to preventing exfiltration of training data by feature reconstruction attacks on model instances trained on the training data during a training job. The system comprises a privacy interface that presents a plurality of modulators for a plurality of training parameters. The modulators are configured to respond to selection commands via the privacy interface to trigger procedural calls. The procedural calls modify corresponding training parameters in the plurality of training parameters for respective training cycles in the training job. The system comprises a trainer configured to execute the training cycles in dependence on the modified training parameters. The trainer can determine a performance accuracy of the model instances for each of the executed training cycles.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to authenticating users using a plurality of non-deterministic registration biometric inputs. During registration, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate sets of feature vectors. The non-deterministic biometric inputs can include a plurality of face images and a plurality of voice samples of a user. A characteristic identity vector for the user can be determined by averaging feature vectors. During authentication, a plurality of non-deterministic biometric inputs are given as input to a trained machine learning model to generate a set of authentication feature vectors. The sets of feature vectors are projected onto a surface of a hyper-sphere. The system can authenticate the user when a cosine distance between the authentication feature vector and a characteristic identity vector for the user is less than a pre-determined threshold.

Abstract: The technology disclosed relates to a system and method for training processing engines. A processing engine can have at least a first processing module and a second processing module. The first processing module in each processing engine is different from a corresponding first processing module in every other processing engine. The second processing module in each processing engine is same as a corresponding second processing module in every other processing engine. The system can include a deployer that deploys each processing engine to a respective hardware module for training. The system can comprise a forward propagator which during forward pass stage can process inputs through first processing modules and produce an intermediate output for each first processing module. The system can comprise a backward propagator which during backward pass stage can determine gradients for each second processing module on corresponding final outputs and ground truths.

Abstract: A computer-implemented method for automatically identifying subjects at risk of Multiple Sclerosis (MS) includes acquiring a plurality of images of a subject's brain using a Magnetic Resonance Imaging (MRI) scanner. A contrast enhancement process is applied to each image to generate a plurality of contrast-enhanced images. An automated lesion detection algorithm is applied to detect one or more lesions present in the contrast-enhanced images. An automated central vein detection algorithm is applied to detect one or more central veins present in the contrast-enhanced images. An automated paramagnetic rim detection algorithm is applied to detect one or more paramagnetic rims present in the contrast-enhanced images. The patient's risk for MS may then be determined based on the one or more of the lesions, central veins, and paramagnetic rims present in the contrast-enhanced images.

Abstract: Systems and methods are provided for providing customer relationship management (“CRM”) data via a messaging interface. In some embodiments, a computing system can be used to generate and provide the messaging interface. The messaging interface can allow a user to provide user input (e.g., text input) and the computing system can access one or more databases that include CRM data in response to receiving the user input. The computing system can also obtain or receive, from the database, a subset of the CRM data relevant to the user input and store the subset of the CRM data. The computing system can also output, via a display device, the subset of the CRM data via the messaging interface. The computing system can also automatically (e.g., without receiving user input) obtain or receive the subset of the CRM data and output the subset of the CRM data via the messaging interface.

Abstract: An autonomous ventilation system includes a variable-speed exhaust fan, a controller, an exhaust hood, and an infrared radiation (“IR”) sensor. The exhaust fan removes air contaminants from an area. The controller is coupled to the exhaust fan and adjusts the speed of the exhaust fan. The exhaust hood is coupled to the exhaust fan and directs air contaminants to the exhaust fan. The IR sensor is coupled to the controller, detects changes in IR index in a zone below the exhaust hood, and communicates information relating to detected changes in IR index to the controller. The controller adjusts the speed of the exhaust fan in response to information relating to detected changes in IR index. The autonomous ventilation system also includes an alignment laser to indicate a point at which the IR sensor is aimed and a field-of-view (“FOV”) indicator to illuminate the zone in which the IR sensor detects changes in IR index.