Qualcomm Incorporated

Gabriele Cesa

Taco Sebastiaan Cohen

Max Welling

Arash Behboodi

Auke Joris Wiggers

Guillaume Konrad Sautiere

Kumar Pratik

Yunfan Zhang

Mirgahney Husham Awadelkareem Mohamed

David Wilson Romero Guzman

Giovanni Luca Marchetti

Certain aspects of the present disclosure provide techniques and apparatus for encoding content using a neural network. An example method generally includes encoding video content into a latent space representation through an encoder implemented by a first machine learning model. A code is generated by upsampling the latent space representation of the video content. A prior is calculated based on a conditional probability of obtaining the upsampled latent space representation conditioned by the latent space representation of the video content. A compressed version of the video content is generated based on a probabilistic model implemented by a second machine learning model, the generated code, and the calculated prior, and the compressed version of the video content is output for transmission.

Alias-free compression of content using artificial neural networks

Certain aspects of the present disclosure provide a method of performing machine learning, comprising: generating a neural network model; and training the neural network model for a task with a first set of input data, wherein: the training uses a total loss functionincluding an equivariance loss componentaccording to=+α, and α>0.

Data and compute efficient equivariant convolutional networks

Certain aspects of the present disclosure provide techniques for wireless communications by an apparatus. Certain techniques include receiving signals corresponding to a MIMO channel matrix; generating a first gram matrix from a basis for a first lattice corresponding to a first signal of the received signals; providing the first gram matrix to a neural lattice reduction model comprising an equivariant neural network configured to generate a current extended Gauss move; generating, with the neural lattice reduction model, a current partial changed basis based on the current extended Gauss move and the basis; executing one or more additional iterations of the neural lattice reduction model; and demapping the MIMO channel matrix based on combining the current partial changed basis and each of the additional partial changed basis generated by each of the one or more additional iterations of the neural lattice reduction model.

Geometric deep learning for lattice reduction

Growing community of inventors

Diemen, Netherlands

Gabriele Cesa