Conventional hardware-friendly quantization methods, such as fixed-point or integer, tend to perform poorly at very low word sizes as their shrinking dynamic ranges cannot adequately capture the wide data distributions commonly seen in sequence transduction models. We present AdaptivFloat, a floating-point inspired number representation format for deep learning that dynamically maximizes and optimally clips its available dynamic range, at a layer granularity, in order to create faithful encoding of neural network parameters. AdaptivFloat consistently produces higher inference accuracies compared to block floating-point, uniform, IEEE-like float or posit encodings at very low precision ($\leq$ 8-bit) across a diverse set of state-of-the-art neural network topologies. And notably, AdaptivFloat is seen surpassing baseline FP32 performance by up to +0.3 in BLEU score and -0.75 in word error rate at weight bit widths that are $\leq$ 8-bit. Experimental results on a deep neural network (DNN) hardware accelerator, exploiting AdaptivFloat logic in its computational datapath, demonstrate per-operation energy and area that is 0.9$\times$ and 1.14$\times$, respectively, that of equivalent bit width integer-based accelerator variants.

Predicting delayed outcomes is an important problem in recommender systems (e.g., if customers will finish reading an ebook). We formalize the problem as an adversarial, delayed online learning problem and consider how a proxy for the delayed outcome (e.g., if customers read a third of the book in 24 hours) can help minimize regret, even though the proxy is not available when making a prediction. Motivated by our regret analysis, we propose two neural network architectures: Factored Forecaster (FF) which is ideal if the proxy is informative of the outcome in hindsight, and Residual Factored Forecaster (RFF) that is robust to a non-informative proxy. Experiments on two real-world datasets for predicting human behavior show that RFF outperforms both FF and a direct forecaster that does not make use of the proxy. Our results suggest that exploiting proxies by factorization is a promising way to mitigate the impact of long delays in human-behavior prediction tasks.

This paper further defines the class of fragment constructive heuristics used to compute feasible solutions for the Traveling Salesman Problem into arc-greedy and node-greedy subclasses. Since these subclasses of heuristics can create subtours, two known methodologies for subtour elimination on symmetric instances are reviewed and are expanded to cover asymmetric problem instances. This paper introduces a third novel methodology, the Greedy Tracker, and compares it to both known methodologies. Computational results are generated across multiple symmetric and asymmetric instances. The results demonstrate the Greedy Tracker is the fastest method for preventing subtours for instances below 400 nodes. A distinction between fragment constructive heuristics and the subtour elimination methodology used to ensure the feasibility of resulting solutions enables the introduction of a new node-greedy fragment heuristic called Ordered Greedy.

Neural Networks (NNs) are steering a new generation of artificial intelligence (AI) applications at the micro-edge. Examples include wireless sensors, wearables and cybernetic systems that collect data and process them to support real-world decisions and controls. For energy autonomy, these applications are typically powered by energy harvesters. As harvesters and other power sources which provide energy autonomy inevitably have power variations, the circuits need to robustly operate over a dynamic power envelope. In other words, the NN hardware needs to be able to function correctly under unpredictable and variable supply voltages. In this paper, we propose a novel NN design approach using the principle of pulse width modulation (PWM). PWM signals represent information with their duty cycle values which may be made independent of the voltages and frequencies of the carrier signals. We design a PWM-based perceptron which can serve as the fundamental building block for NNs, by using an entirely new method of realising arithmetic in the PWM domain. We analyse the proposed approach building from a 3x3 perceptron circuit to a complex multi-layer NN. Using handwritten character recognition as an exemplar of AI applications, we demonstrate the power elasticity, resilience and efficiency of the proposed NN design in the presence of functional and parametric variations including large voltage variations in the power supply.

Parallel exploration is a key to a successful search. The recently proposed Negatively Correlated Search (NCS) achieved this ability by constructing a set of negatively correlated search processes and has been applied to many real-world problems. In NCS, the key technique is to explicitly model and maximize the diversity among search processes in parallel. However, the original diversity model was mostly devised by intuition, which introduced several drawbacks to NCS. In this paper, a mathematically principled diversity model is proposed to solve the existing drawbacks of NCS, resulting a new NCS framework. A new instantiation of NCS is also derived and its effectiveness is verified on a set of multi-modal continuous optimization problems.

Palo Alto, CA, 94306, USA ABSTRACT We propose an architecture to jointly learn word and label embeddings for slot filling in spoken language understanding. The proposed approach encodes labels using a combination of word embeddings and straightforward word-label association from the training data. Compared to the state-of- the-art methods, our approach does not require label embed-dings as part of the input and therefore lends itself nicely to a wide range of model architectures. In addition, our architecture computes contextual distances between words and labels to avoid adding contextual windows, thus reducing memory footprint. We validate the approach on established spoken dialogue datasets and show that it can achieve state-of-the-art performance with much fewer trainable parameters. Index T erms-- Slot-filling, recurrent neural network, distributional semantics, sequence labelling 1. INTRODUCTION In spoken language understanding (SLU), an essential step is to associate each word in an utterance with one semantic class label. These annotated utterances can then serve as a basis for higher level SLU tasks, such as topic identification and dialogue response generation. This process of semantic label tagging in SLU, dubbed slot filling, labels utterance sequences with tags under a specific scheme. As an example, the BIO scheme prefixes tags with one of the characters { B, I, O } to indicate the continuity of a tag: Begin, Inside, or Outside, e.g., B-price indicates this position is the beginning of the tag price. Researchers also developed deep learning architecture for slot filling, e.g., [1, 2, 3].

We demonstrate that models trained only in simulation can be used to solve a manipulation problem of unprecedented complexity on a real robot. This is made possible by two key components: a novel algorithm, which we call automatic domain randomization (ADR) and a robot platform built for machine learning. ADR automatically generates a distribution over randomized environments of ever-increasing difficulty. Control policies and vision state estimators trained with ADR exhibit vastly improved sim2real transfer. For control policies, memory-augmented models trained on an ADR-generated distribution of environments show clear signs of emergent meta-learning at test time. The combination of ADR with our custom robot platform allows us to solve a Rubik's cube with a humanoid robot hand, which involves both control and state estimation problems. Videos summarizing our results are available: https://openai.com/blog/solving-rubiks-cube/

One of the key challenges in the semantic mapping problem in postdisaster environments is how to analyze a large amount of data efficiently with minimal supervision. To address this challenge, we propose a deep learning-based semantic mapping tool consisting of three main ideas. First, we develop a frugal semantic segmentation algorithm that uses only a small amount of labeled data. Next, we investigate on the problem of learning to detect a new class of object using just a few training examples. Finally, we develop an explainable cost map learning algorithm that can be quickly trained to generate traversability cost maps using only raw sensor data such as aerial-view imagery. This paper presents an overview of the proposed idea and the lessons learned.

From its inception, AI has had a rather ambivalent relationship to humans---swinging between their augmentation and replacement. Now, as AI technologies enter our everyday lives at an ever increasing pace, there is a greater need for AI systems to work synergistically with humans. To do this effectively, AI systems must pay more attention to aspects of intelligence that helped humans work with each other---including social intelligence. I will discuss the research challenges in designing such human-aware AI systems, including modeling the mental states of humans in the loop, recognizing their desires and intentions, providing proactive support, exhibiting explicable behavior, giving cogent explanations on demand, and engendering trust. I will survey the progress made so far on these challenges, and highlight some promising directions. I will also touch on the additional ethical quandaries that such systems pose. I will end by arguing that the quest for human-aware AI systems broadens the scope of AI enterprise, necessitates and facilitates true inter-disciplinary collaborations, and can go a long way towards increasing public acceptance of AI technologies.

Model-free reinforcement learning is known to be memory and computation efficient and more amendable to large scale problems. In this paper, two model-free algorithms are introduced for learning infinite-horizon average-reward Markov Decision Processes (MDPs). The first algorithm reduces the problem to the discounted-reward version and achieves $\mathcal{O}(T^{2/3})$ regret after $T$ steps, under the minimal assumption of weakly communicating MDPs. The second algorithm makes use of recent advances in adaptive algorithms for adversarial multi-armed bandits and improves the regret to $\mathcal{O}(\sqrt{T})$, albeit with a stronger ergodic assumption. To the best of our knowledge, these are the first model-free algorithms with sub-linear regret (that is polynomial in all parameters) in the infinite-horizon average-reward setting.