A new DeepMind paper introduces two architectures designed for the efficient use of Tensor Processing Units (TPUs) in reinforcement learning (RL) research at scale.

Deep learning (DL) frameworks such as TensorFlow, PyTorch and JAX enable easy, rapid model prototyping while also optimizing execution speed for model training. Although such frameworks are popular across the general DL community, scalable research platforms for deep RL remain relatively underdeveloped.

The proposed DeepMind architectures, Anakin and Sebulba, address this deficiency, demonstrating how TPU-based RL platforms can deliver exceptional performance at low cost.


Spiking Neural Networks (SNN) represent the third generation of artificial neural networks. SNN models are built using both spatial and temporal aspects of the input data, and as such advance a step closer to true brain-inspired processing. SNNs have shown great promise and potential in low-power sensory-processing and edge computing hardware platforms.

In the paper An Error-Propagation Spiking Neural Network Compatible With Neuromorphic Processors, researchers from ETH Zurich leverage existing spike-based learning circuits to propose a biologically plausible architecture that is highly successful in classifying distinct complex spatio-temporal spike patterns.

The study advances the design and development of ultra-low-power mixed-signal…


Large-scale transformer-based language models have produced substantial gains in the field of natural language processing (NLP). Training such models however is challenging, for two reasons: No single GPU has enough memory to accommodate parameter totals which have grown exponentially in recent years, and even if there were a way to train these parameters on single GPU, limited computing power would result in unrealistically long training times without model parallelism.

In the paper Efficient Large-Scale Language Model Training on GPU Clusters, a research team from NVIDIA, Stanford University and Microsoft Research propose a novel parallelization schedule which improves throughput by more…


In the field of reinforcement learning (RL), task specifications are typically designed by experts. Learning from demonstrations and preferences requires a great deal of human interaction, and hand-coded reward functions are notoriously difficult to specify. If all these hand-designed RL system parts and specifications could be replaced with automatically learned components — as is increasingly the case in other AI areas — that would be a huge breakthrough.

In a new paper, a research team from ETH Zurich and UC Berkeley propose Deep Reward Learning by Simulating the Past (Deep RLSP), a novel algorithm that represents rewards directly as a…


A new Google Brain and New York University study argues that the current evaluation techniques for natural language understanding (NLU) tasks are broken, and proposes guidelines designed to produce better NLU benchmarks.

Contemporary NLU studies tend to focus on improving results on benchmark datasets that feature roughly independent and identically distributed (IID) training, evaluating and testing. The researchers however say such benchmark-driven NLU research has become problematic, as “unreliable and biased systems score so highly on standard benchmarks that there is little room for researchers who develop better systems to demonstrate their improvements.”

A recent trend to address this issue…


Knowledge graphs (KGs) are graphs used to accumulate and convey real-world knowledge. KG nodes capture information about entities of interest (like people, places or events) in a given domain or task, while the edges represent the connections between them. To provide vital information for related tasks such as Knowledge Base Question Answering (KBQA), various semantic web technologies have been employed to represent KGs with explicit semantics, defining a type for each node. A “Tylor Swift” node for example could be classified as a “popular singer” type. …


OpenAI’s powerful GPT-3 large language model was a gamechanger in the machine learning community, and numerous illustrative demos have emerged since its June 2020 release. Debuild founder Sharif Shameem’s demo shows how GPT-3 lets users describe a desired layout in plain language, then sit back while the generator produces the appropriate JSX code.

The transformer-based GPT-3 has also proven a powerful solution across a variety of natural language processing (NLP) tasks, indicating the huge potential for NLP applications to process source code and crack software engineering tasks. This promising direction has however remained relatively under-explored. Until now.

In the…


Albert Einstein once said that “wisdom is not a product of schooling, but the lifelong attempt to acquire it.” Centuries of human progress have been built on our brains’ ability to continually acquire, fine-tune and transfer knowledge and skills. Such continual learning however remains a long-standing challenge in machine learning (ML), where the ongoing acquisition of incrementally available information from non-stationary data often leads to catastrophic forgetting problems.

Gradient-based deep architectures have spurred the development of continual learning in recent years, but continual learning algorithms are often designed and implemented from scratch with different assumptions, settings, and benchmarks, making them…


Powerful transformer models have been widely used in autoregressive generation, where they have advanced the state-of-the-art beyond recurrent neural networks (RNNs). However, because the output words for these models are incrementally predicted conditioned on the prefix, the generation requires quadratic time complexity with regard to sequence length.

As the performance of transformer models increasingly relies on large-scale pretrained transformers, this long sequence generation issue has become increasingly problematic. To address this, a research team from the University of Washington, Microsoft, DeepMind and Allen Institute for AI have developed a method to convert a pretrained transformer into an efficient RNN. …


Recent breakthroughs in machine learning (ML) have enabled AI systems to assume increasingly important roles in real-world decision-making. Studies have suggested however that such systems may be prone to biases that could result in discrimination against individuals on the basis of racial and gender characteristics.

In the 2011 paper Fairness Through Awareness, Cynthia Dwork et al. propose that an ML model lacks “individual fairness” if a pair of valid inputs which are otherwise close to each other (according to an appropriate metric) are treated differently by the model (different class label, or a large difference in output). …

Synced

AI Technology & Industry Review — syncedreview.com | Newsletter: http://bit.ly/2IYL6Y2 | Share My Research http://bit.ly/2TrUPMI | Twitter: @Synced_Global

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store