Graph neural networks (GNNs) have shown high potential for a variety of real-world, challenging applications, but one of the major obstacles in GNN research is the lack of large-scale flexible datasets. Most existing public datasets for GNNs are relatively small, which limits the ability of GNNs to generalize to unseen data. The few existing large-scale graph datasets provide very limited labeled data. This makes it difficult to determine if the GNN model's low accuracy for unseen data is inherently due to insufficient training data or if the model failed to generalize.

We propose Monte Carlo methods to leverage both sparsity and quantization to compress gradients of neural networks throughout training. On top of reducing the communication exchanged between multiple workers in a distributed setting, we also improve the computational efficiency of each worker. Our method, called Monte Carlo Gradient Quantization (MCGQ), shows faster convergence and higher performance than existing quantization methods on image classification and language modeling.

We propose Monte Carlo methods to leverage both sparsity and quantization to compress gradients of neural networks throughout training. On top of reducing the communication exchanged between multiple workers in a distributed setting, we also improve the computational efficiency of each worker. Our method, called Monte Carlo Gradient Quantization (MCGQ), shows faster convergence and higher performance than existing quantization methods on image classification and language modeling.

We demonstrate that 1x1-convolutions in 1D time-channel separable convolutions may be replaced by constant, sparse random ternary matrices with weights in {−1, 0, +1}. Such layers do not perform any multiplications and do not require training. Moreover, the matrices may be generated on the chip during computation and therefore do not require any memory access. With the same parameter budget, we can afford deeper and more expressive models, improving the Pareto frontiers of existing models on several tasks.

We have implemented a recently proposed partially linear multiuser detection algorithm in reproducing kernel Hilbert spaces (RKHSs) on a GPU-accelerated platform. Our proof of concept combines the robustness of linear detection and non-linear detection for the non-orthogonal multiple access (NOMA) based massive connectivity scenario. Mastering the computation of the vast number of inner products (which involve kernel evaluations) is a challenge in ultra-low latency (ULL) applications due to the sub-millisecond latency requirement.

Abstract Artificial neural networks can be represented by paths. Generated as random walks on a dense network graph, we find that the resulting sparse networks allow for deterministic initialization and even weights with fixed sign. Such networks can be trained sparse from scratch, avoiding the expensive procedure of training a dense network and compressing it afterwards. Although sparse, weights are accessed as contiguous blocks of memory.

We investigate adaptive projected sub gradient method (APSM) and neural network (NN) machine learning techniques to address the challenge of digital self-interference cancellation in full-duplex communications. To this end, we compare both approaches in terms of their interference suppression capabilities, their computational complexity, and discuss their potential of continual training. Both approaches can take advantage of massively parallel processing in the digital domain, resulting in a significantly reduced end-to-end latency.

We revisit the design and retraining capabilities of neural network (NN)-based orthogonal frequency division multiplex (OFDM) receivers that combine channel estimation, equalization and soft-demapping for time-varying and frequency selective wireless channels. Attracted by the inherent advantages of small NNs in terms of computational complexity during inference and (re-)training, we first analyze the performance of different neural receiver architectures, including versions with reduced complexity.