Abstract: A method for training a generator, by a generator training system including a processor and memory, includes: extracting training statistical characteristics from a batch normalization layer of a pre-trained model, the training statistical characteristics including a training mean ? and a training variance ?2; initializing a generator configured with generator parameters; generating a batch of synthetic data using the generator; supplying the batch of synthetic data to the pre-trained model; measuring statistical characteristics of activations at the batch normalization layer and at the output of the pre-trained model in response to the batch of synthetic data, the statistical characteristics including a measured mean, and {circumflex over (?)}? measured variance {circumflex over (?)}?2; computing a training loss in accordance with a loss function L? based on ?, ?2, {circumflex over (?)}?, and {circumflex over (?)}?2; and iteratively updating the generator parameters in accordance with the training loss until

Abstract: A system and a method for performing gesture recognition are disclosed, the method comprising detecting a gesture using a primary modality; evaluating an expected accuracy gain (EAG) to identify a modality that yields a maximum relative EAG among the primary modality and one or more secondary modalities; and activating the one or more secondary modalities for detecting the gesture if the one or more secondary modalities correspond to the modality that yields the maximum relative EAG.

Abstract: A convolutional neural network (CNN) system for generating a classification for an input image is presented. The CNN system comprises circuitry running on clock cycles and configured to compute a product of two received values, and at least one non-transitory computer-readable medium that stores instructions for the circuitry to derive a feature map based on at least the input image; puncture at least one selection among the feature map and a kernel by setting the value of an element at an index of the at least one selection to zero and cyclic shifting a puncture pattern to achieve a 1/d reduction in number of clock cycles, where d is an integer and puncture interval value >1. The feature map is convolved with the kernel to generate an output, and a classification of the input image is generated based on the output.

Abstract: A method and system for training a neural network are provided. The method includes receiving an input image, selecting at least one data augmentation method from a pool of data augmentation methods, generating an augmented image by applying the selected at least one data augmentation method to the input image, and generating a mixed image from the input image and the augmented image.

Abstract: Methods and systems for performing video prediction, including obtaining an input frame from among a plurality of frames included in an input video; extracting a first feature map by providing the input frame to a first plurality of feature extraction layers and a first strided convolutional layer included in an encoder; providing the first feature map and at least one neighboring first feature map corresponding to at least one neighboring frame to a first fusion module included in the encoder; fusing the first feature map with the at least one neighboring first feature map to generate a fused first feature map using the first fusion module; generating a prediction corresponding to the input frame based on the fused first feature map using a decoder; and performing a video prediction task using the prediction.

Abstract: Methods and devices are provided for performing federated learning. A global model is distributed from a server to a plurality of client devices. At each of the plurality of client devices: model inversion is performed on the global model to generate synthetic data; the global model is on an augmented dataset of collected data and the synthetic data to generate a respective client model; and the respective client model is transmitted to the server. At the server: client models are received from the plurality of client devices, where each client model is received from a respective client device of the plurality of client devices; model inversion is performed on each client model to generate a synthetic dataset; the client models are averaged to generate an averaged model; and the averaged model is trained using the synthetic dataset to generate an updated model.

Abstract: A system and a method are disclosed for hardware-aware pruning of conformer networks. In some embodiments, the method includes: training a neural network, the training including: performing a first pruning operation, on the neural network, after a first training epoch, and performing a second pruning operation, on the neural network, after a second training epoch and after the first pruning operation, wherein each of the pruning operations results in a respective pruning fraction, the respective pruning fraction being a function of an index of a training epoch preceding the pruning operation.

Abstract: Apparatuses and methods of manufacturing same, systems, and methods for performing network parameter quantization in deep neural networks are described. In one aspect, multi-dimensional vectors representing network parameters are constructed from a trained neural network model. The multi-dimensional vectors are quantized to obtain shared quantized vectors as cluster centers, which are fine-tuned. The fine-tuned and shared quantized vectors/cluster centers are then encoded. Decoding reverses the process.

Abstract: A system for performing echo cancellation includes: a processor configured to: receive a far-end signal; record a microphone signal including: a near-end signal; and an echo signal corresponding to the far-end signal; extract far-end features from the far-end signal; extract microphone features from the microphone signal; compute estimated near-end features by supplying the microphone features and the far-end features to an acoustic echo cancellation module including a recurrent neural network including: an encoder including a plurality of gated recurrent units; and a decoder including a plurality of gated recurrent units; compute an estimated near-end signal from the estimated near-end features; and transmit the estimated near-end signal to the far-end device. The recurrent neural network may include a contextual attention module; and the recurrent neural network may take, as input, a plurality of error features computed based on the far-end features, the microphone features, and acoustic path parameters.

Abstract: A method for training a generator, by a generator training system including a processor and memory, includes: extracting training statistical characteristics from a batch normalization layer of a pre-trained model, the training statistical characteristics including a training mean ? and a training variance ?2; initializing a generator configured with generator parameters; generating a batch of synthetic data using the generator; supplying the batch of synthetic data to the pre-trained model; measuring statistical characteristics of activations at the batch normalization layer and at the output of the pre-trained model in response to the batch of synthetic data, the statistical characteristics including a measured mean {circumflex over (?)}? and a measured variance {circumflex over (?)}?2; computing a training loss in accordance with a loss function L? based on ?, ?2, {circumflex over (?)}?, and {circumflex over (?)}?2; and iteratively updating the generator parameters in accordance with the training loss unti

Abstract: A system and a method are disclosed, the method including receiving, by a first local controller of a first edge device, an input associated with an environment in which the first edge device operates, using a first machine-learning algorithm, determining, by the first local controller, a parameter for a pre-trained modem algorithm of the first edge device based on the input, executing a task on the first edge device based on executing the pre-trained modem algorithm with the parameter, determining a result of executing the task, training the first machine-learning algorithm, generating a first update to the first machine-learning algorithm based on the training, sending the first update to a server, receiving, from the server, a server update to the first machine-learning algorithm, and based on the server update, updating the first machine-learning algorithm.

Abstract: A system and a method are disclosed for tuning parameters of a large language model. The method comprises identifying first weights of a machine learning (ML) model. Second weights are received from a client device. The second weights may be based on updating, by the client device, the first weights. An update matrix may be generated based on the second weights. The update matrix may be decomposed into first decomposition matrices. Singular values that satisfy a criterion may be identified based on the first decomposition matrices. Singular vectors may be identified based on the singular values. Second decomposition matrices may be identified based on the singular vectors. Updates may be received from the client device of third weights associated with the second decomposition matrices. An updated ML model may be generated based on the updates of the third weights. An inference may be generated based on the updated ML model.

Abstract: A system including: one or more processors; and memory including instructions that, when executed by the one or more processors, cause the one or more processors to: generate augmented input data by mixing noise components of training data; train a first neural network based on the augmented input data and ground truth data of the training data to output a first prediction of clean speech; lock trainable parameters of the first neural network as a result of the training of the first neural network; and train a second neural network according to the augmented input data and predictions of the first neural network to output a second prediction of the clean speech.

Abstract: A method of depth detection based on a plurality of video frames includes receiving a plurality of input frames including a first input frame, a second input frame, and a third input frame respectively corresponding to different capture times, convolving the first to third input frames to generate a first feature map, a second feature map, and a third feature map corresponding to the different capture times, calculating a temporal attention map based on the first to third feature maps, the temporal attention map including a plurality of weights corresponding to different pairs of feature maps from among the first to third feature maps, each weight of the plurality of weights indicating a similarity level of a corresponding pair of feature maps, and applying the temporal attention map to the first to third feature maps to generate a feature map with temporal attention.

Abstract: A method and system for providing Gaussian weighted self-attention for speech enhancement are herein provided. According to one embodiment, the method includes receiving an input noise signal, generating a score matrix based on the received input noise signal, and applying a Gaussian weighted function to the generated score matrix.

Abstract: A system for performing echo cancellation includes: a processor configured to: receive a far-end signal; record a microphone signal including: a near-end signal; and an echo signal corresponding to the far-end signal; extract far-end features from the far-end signal; extract microphone features from the microphone signal; compute estimated near-end features by supplying the microphone features and the far-end features to an acoustic echo cancellation module including a recurrent neural network including: an encoder including a plurality of gated recurrent units; and a decoder including a plurality of gated recurrent units; compute an estimated near-end signal from the estimated near-end features; and transmit the estimated near-end signal to the far-end device. The recurrent neural network may include a contextual attention module; and the recurrent neural network may take, as input, a plurality of error features computed based on the far-end features, the microphone features, and acoustic path parameters.

Abstract: A system and a method to train a neural network are disclosed. A first image is weakly and strongly augmented. The first image, the weakly and strongly augmented first images are input into a feature extractor to obtain augmented features. Each weakly augmented first image is input to a corresponding first expert head to determine a supervised loss for each weakly augmented first image. Each strongly augmented first image is input to a corresponding second expert head to determine a diversity loss for each strongly augmented first image. The feature extractor is trained to minimize the supervised loss on weakly augmented first images and to minimize a multi-expert consensus loss on strongly augmented first images. Each first expert head is trained to minimize the supervised loss for each weakly augmented first image, and each second expert head is trained to minimize the diversity loss for each strongly augmented first image.

Abstract: A system for performing echo cancellation includes: a processor configured to: receive a far-end signal; record a microphone signal including: a near-end signal; and an echo signal corresponding to the far-end signal; extract far-end features from the far-end signal; extract microphone features from the microphone signal; compute estimated near-end features by supplying the microphone features and the far-end features to an acoustic echo cancellation module including a recurrent neural network including: an encoder including a plurality of gated recurrent units; and a decoder including a plurality of gated recurrent units; compute an estimated near-end signal from the estimated near-end features; and transmit the estimated near-end signal to the far-end device. The recurrent neural network may include a contextual attention module; and the recurrent neural network may take, as input, a plurality of error features computed based on the far-end features, the microphone features, and acoustic path parameters.

Abstract: A computer vision (CV) training system, includes: a supervised learning system to estimate a supervision output from one or more input images according to a target CV application, and to determine a supervised loss according to the supervision output and a ground-truth of the supervision output; an unsupervised learning system to determine an unsupervised loss according to the supervision output and the one or more input images; a weakly supervised learning system to determine a weakly supervised loss according to the supervision output and a weak label corresponding to the one or more input images; and a joint optimizer to concurrently optimize the supervised loss, the unsupervised loss, and the weakly supervised loss.

Abstract: Disclosed is a method including receiving, in a semantic segmentation network, input data from a plurality of frames, computing a ground truth label on the plurality of frames, generating a ground truth temporal semantic boundary map from the ground truth label on the plurality of frames, generating a predicted temporal semantic boundary map based on an output of the input data, and determining a loss based on the ground truth temporal semantic boundary map and the predicted temporal semantic boundary map.