Lawrence Livermore National Security, LLC

Arizona State University

Jayaraman Jayaraman Thiagarajan

Andreas Spanias

Peer-Timo Bremer

Huan Song

Rushil Anirudh

Karthikeyan Natesan Ramamurthy

Hyojin Kim

David H Frakes

Vivek Sivaraman Narayanaswamy

Uday Shanthamallu

Visar Berisha

Sameeksha Katoch

Bhavya Kailkhura

Deepta Rajan

Megan Willi

Prasanna Sattigeri

Various embodiments of a system and methods for audio source separation via multi-scale feature learning are disclosed.

Systems and methods for audio source separation via multi-scale feature learning

Various embodiments of a system and associated method for audio source separation based on generative priors trained on individual sources. Through the use of projected gradient descent optimization, the present approach simultaneously searches in the source-specific latent spaces to effectively recover the constituent sources. Though the generative priors can be defined in the time domain directly, it was found that using spectral domain loss functions leads to good-quality source estimates.

Systems and methods for unsupervised audio source separation using generative priors

Various embodiments of systems and methods for attention models with random features for multi-layered graph embeddings are disclosed.

Systems and methods for a supra-graph based attention model for multi-layered embeddings and deep learning applications

A system for classifying a target image with segments having attributes is provided. The system generates a graph for the target image that includes vertices representing segments of the image and edges representing relationships between the connected vertices. For each vertex, the system generates a subgraph that includes the vertex as a home vertex and neighboring vertices representing segments of the target image within a neighborhood of the segment represented by the home vertex. The system applies an autoencoder to each subgraph to generate latent variables to represent the subgraph. The system applies a machine learning algorithm to a feature vector comprising a universal image representation of the target image that is derived from the generated latent variables of the subgraphs to generate a classification for the target image.

Universal image representation based on a multimodal graph

A system for time series analysis using attention models is disclosed. The system may capture dependencies across different variables through input embedding and may map the order of a sample appearance to a randomized lookup table via positional encoding. The system may capture capturing dependencies within a single sequence through a self-attention mechanism and determine a range of dependency to consider for each position being analyzed. The system may obtain an attention weighting to other positions in the sequence through computation of an inner product and utilize the attention weighting to acquire a vector representation for a position and mask the sequence to enable causality. The system may employ a dense interpolation technique for encoding partial temporal ordering to obtain a single vector representation and a linear layer to obtain logits from the single vector representation. The system may use a type dependent final prediction layer.

Systems and methods for time series analysis using attention models

A method including receiving an input data set. The input data set can include one of a feature domain set or a kernel matrix. The method also can include constructing dense embeddings using: (i) Nystr&#xf6;m approximations on the input data set when the input data set comprises the kernel matrix, and (ii) clustered Nystr&#xf6;m approximations on the input data set when the input data set comprises the feature domain set. The method additionally can include performing representation learning on each of the dense embeddings using a multi-layer fully-connected network for each of the dense embeddings to generate latent representations corresponding to each of the dense embeddings. The method further can include applying a fusion layer to the latent representations corresponding to the dense embeddings to generate a combined representation. The method additionally can include performing classification on the combined representation. Other embodiments of related systems and methods are also disclosed.

Systems and methods for customizing kernel machines with deep neural networks

Systems and methods for a supra-fusion graph attention model for multi-layered embeddings and deep learning applications

Various embodiments of a systems and methods for a triplet network having speaker diarization are disclosed.

Systems and methods for a triplet network with attention for speaker diartzation

Methods and systems are provided to generate an uncorrupted version of an image given an observed image that is a corrupted version of the image. In some embodiments, a corruption mimicking ('CM') system iteratively trains a corruption mimicking network ('CMN') to generate corrupted images given modeled images, updates latent vectors based on differences between the corrupted images and observed images, and applies a generator to the latent vectors to generate modeled images. The training, updating, and applying are performed until modeled images that are input to the CMN result in corrupted images that approximate the observed images. Because the CMN is trained to mimic the corruption of the observed images, the final modeled images represented the uncorrupted version of the observed images.

Mimicking of corruption in images

A system for identifying in an image an object that is commonly found in a collection of images and for identifying a portion of an image that represents an object based on a consensus analysis of segmentations of the image. The system collects images of containers that contain objects for generating a collection of common objects within the containers. To process the images, the system generates a segmentation of each image. The image analysis system may also generate multiple segmentations for each image by introducing variations in the selection of voxels to be merged into a segment. The system then generates clusters of the segments based on similarity among the segments. Each cluster represents a common object found in the containers. Once the clustering is complete, the system may be used to identify common objects in images of new containers based on similarity between segments of images and the clusters.

Identification of uncommon objects in containers

A system for identifying objects in an image is provided. The system identifies segments of an image that may contain objects. For each segment, the system generates a segment score by inputting to a multi-scale neural network windows of multiple scales that include the segment that have been resampled to a fixed window size. A multi-scale neural network includes a feature extracting convolutional neural network ('feCNN') for each scale and a classifier that inputs each feature of each feCNN. The segment score indicates whether the segment contains an object. The system generates a pixel score for pixels of the image. The pixel score for a pixel indicates that that pixel is within an object based on the segment scores of segments that contain that pixel. The system then identifies the object based on the pixel scores of neighboring pixels.

Multi-scale deep learning system

Methods and systems for recovering corrupted/degraded images using approximations obtained from an ensemble of multiple sparse models are disclosed. Sparse models may represent images parsimoniously using elementary patterns from a 'dictionary' matrix. Various embodiments of the present disclosure involve simple and computationally efficient dictionary design approach along with low-complexity reconstruction procedure that may use a parallel-friendly table-lookup process. Multiple dictionaries in an ensemble model may be inferred sequentially using greedy forward-selection approach and can incorporate bagging/boosting strategies, taking into account application-specific degradation. Recovery performance obtained using the proposed approaches with image super resolution and compressive recovery can be comparable to or better than existing sparse modeling based approaches, at reduced computational complexity. By including ensemble models in hierarchical multilevel learning, where multiple dictionaries are inferred in each level, further performance improvements can be obtained in image recovery, without significant increase in computational complexity.

Ensemble sparse models for image analysis and restoration

Measuring the number of glomeruli in the entire, intact kidney using non-destructive techniques is of immense importance in studying several renal and systemic diseases. In particular, a recent Magnetic Resonance Imaging (MRI) technique, based on injection of a contrast agent, cationic ferritin, has been effective in identifying glomerular regions in the kidney. In various embodiments, a low-complexity, high accuracy method for obtaining the glomerular count from such kidney MRI images is described. This method employs a patch-based approach for identifying a low-dimensional embedding that enables the separation of glomeruli regions from the rest. By using only a few images marked by the expert for learning the model, the method provides an accurate estimate of the glomerular number for any kidney image obtained with the contrast agent. In addition, the implementation of our method shows that this method is near real-time, and can process about 5 images per second.

Growing community of inventors

Milpitas, CA, United States of America

Jayaraman Jayaraman Thiagarajan