Recently, deep learning has been successfully applied to a variety of networking problems. A fundamental challenge is that when the operational environment for a learning-augmented system differs from its training environment, such systems often make badly informed decisions, leading to bad performance. We argue that safely deploying learning-driven systems requires being able to determine, in real time, whether system behavior is coherent, for the purpose of defaulting to a reasonable heuristic when this is not so. We term this the online safety assurance problem (OSAP). We present three approaches to quantifying decision uncertainty that differ in terms of the signal used to infer uncertainty. We illustrate the usefulness of online safety assurance in the context of the proposed deep reinforcement learning (RL) approach to video streaming. While deep RL for video streaming bests other approaches when the operational and training environments match, it is dominated by simple heuristics when the two differ. Our preliminary findings suggest that transitioning to a default policy when decision uncertainty is detected is key to enjoying the performance benefits afforded by leveraging ML without compromising on safety.

Virtual Assistants can be quite literal at times. If the user says "tell Bob I love him," most virtual assistants will extract the message "I love him" and send it to the user's contact named Bob, rather than properly converting the message to "I love you." We designed a system to allow virtual assistants to take a voice message from one user, convert the point of view of the message, and then deliver the result to its target user. We developed a rule-based model, which integrates a linear text classification model, part-of-speech tagging, and constituency parsing with rule-based transformation methods. We also investigated Neural Machine Translation (NMT) approaches, including LSTMs, CopyNet, and T5. We explored 5 metrics to gauge both naturalness and faithfulness automatically, and we chose to use BLEU plus METEOR for faithfulness and relative perplexity using a separately trained language model (GPT) for naturalness. Transformer-Copynet and T5 performed similarly on faithfulness metrics, with T5 achieving slight edge, a BLEU score of 63.8 and a METEOR score of 83.0. CopyNet was the most natural, with a relative perplexity of 1.59. CopyNet also has 37 times fewer parameters than T5. We have publicly released our dataset, which is composed of 46,565 crowd-sourced samples.

The resizing of images, which is typically a required part of preprocessing for computer vision systems, is vulnerable to attack. Images can be created such that the image is completely different at machine-vision scales than at other scales and the default settings for some common computer vision and machine learning systems are vulnerable. We show that defenses exist and are trivial to administer provided that defenders are aware of the threat. These attacks and defenses help to establish the role of input sanitization in machine learning.

Network binarization is a promising hardware-aware direction for creating efficient deep models. Despite its memory and computational advantages, reducing the accuracy gap between such models and their real-valued counterparts remains an unsolved challenging research problem. To this end, we make the following 3 contributions: (a) To increase model capacity, we propose Expert Binary Convolution, which, for the first time, tailors conditional computing to binary networks by learning to select one data-specific expert binary filter at a time conditioned on input features. Overall, our method improves upon prior work, with no increase in computational cost by 6%, reaching a groundbreaking 71% on ImageNet classification. A promising, hardware-aware, direction for designing efficient deep learning models case is that of network binarization, in which filter and activation values are restricted to two states only: 1 [36; 11]. This comes with two important advantages: (a) it compresses the weights by a factor of 32 via bit-packing, and (b) it replaces the computationally expensive multiply-add with bit-wise xnor and popcount operations, offering in practice a speedup of 58 on a CPU [36]. Despite this, how to reduce the accuracy gap between a binary model and its real-valued counterpart remains an open problem and it is currently the major impediment for their wide scale adoption. In this work, we propose to approach this challenging problem from 3 key perspectives: 1. Model capacity: To increase model capacity, we firstly introduce the first application of Conditional Computing [3; 2; 47] to the case of a binary networks, which we call Expert Binary Convolution. For each convolutional layer, rather than learning a weight tensor that is expected to generalize well across the entire input space, we learn a set of N experts each of which is tuned to specialize to portions of it.

Deep neural networks can achieve great successes when presented with large data sets and sufficient computational resources. However, their ability to learn new concepts quickly is quite limited. Meta-learning is one approach to address this issue, by enabling the network to learn how to learn. The exciting field of Deep Meta-Learning advances at great speed, but lacks a unified, insightful overview of current techniques. This work presents just that. After providing the reader with a theoretical foundation, we investigate and summarize key methods, which are categorized into i) metric-, ii) model-, and iii) optimization-based techniques. In addition, we identify the main open challenges, such as performance evaluations on heterogeneous benchmarks, and reduction of the computational costs of meta-learning.

For many reasoning-heavy tasks, it is challenging to find an appropriate end-to-end differentiable approximation to domain-specific inference mechanisms. Neural-Symbolic (NeSy) AI divides the end-to-end pipeline into neural perception and symbolic reasoning, which can directly exploit general domain knowledge such as algorithms and logic rules. However, it suffers from the exponential computational complexity caused by the interface between the two components, where the neural model lacks direct supervision, and the symbolic model lacks accurate input facts. As a result, they usually focus on learning the neural model with a sound and complete symbolic knowledge base while avoiding a crucial problem: where does the knowledge come from? In this paper, we present Abductive Meta-Interpretive Learning ($Meta_{Abd}$), which unites abduction and induction to learn perceptual neural network and first-order logic theories simultaneously from raw data. Given the same amount of domain knowledge, we demonstrate that $Meta_{Abd}$ not only outperforms the compared end-to-end models in predictive accuracy and data efficiency but also induces logic programs that can be re-used as background knowledge in subsequent learning tasks. To the best of our knowledge, $Meta_{Abd}$ is the first system that can jointly learn neural networks and recursive first-order logic theories with predicate invention.

We present a method for learning representations of entities, that uses a Transformer-based architecture as an entity encoder, and link prediction training on a knowledge graph with textual entity descriptions. We demonstrate that our approach can be applied effectively for link prediction in different inductive settings involving entities not seen during training, outperforming related state-of-the-art methods (22% MRR improvement on average). We provide evidence that the learned representations transfer to other tasks that do not require fine-tuning the entity encoder. In an entity classification task we obtain an average improvement of 16% accuracy compared with baselines that also employ pre-trained models. For an information retrieval task, significant improvements of up to 8.8% in NDCG@10 were obtained for natural language queries.

In the context of chit-chat dialogues it has been shown that endowing systems with a persona profile is important to produce more coherent and meaningful conversations. Still, the representation of such personas has thus far been limited to a fact-based representation (e.g. "I have two cats."). We argue that these representations remain superficial w.r.t. the complexity of human personality. In this work, we propose to make a step forward and investigate stance-based persona, trying to grasp more profound characteristics, such as opinions, values, and beliefs to drive language generation. To this end, we introduce a novel dataset allowing to explore different stance-based persona representations and their impact on claim generation, showing that they are able to grasp abstract and profound aspects of the author persona.

Past work on story generation has demonstrated the usefulness of conditioning on a generation plan to generate coherent stories. However, these approaches have used heuristics or off-the-shelf models to first tag training stories with the desired type of plan, and then train generation models in a supervised fashion. In this paper, we propose a deep latent variable model that first samples a sequence of anchor words, one per sentence in the story, as part of its generative process. During training, our model treats the sequence of anchor words as a latent variable and attempts to induce anchoring sequences that help guide generation in an unsupervised fashion. We conduct experiments with several types of sentence decoder distributions: left-to-right and non-monotonic, with different degrees of restriction. Further, since we use amortized variational inference to train our model, we introduce two corresponding types of inference network for predicting the posterior on anchor words. We conduct human evaluations which demonstrate that the stories produced by our model are rated better in comparison with baselines which do not consider story plans, and are similar or better in quality relative to baselines which use external supervision for plans. Additionally, the proposed model gets favorable scores when evaluated on perplexity, diversity, and control of story via discrete plan.

It is challenging to automatically evaluate the answer of a QA model at inference time. Although many models provide confidence scores, and simple heuristics can go a long way towards indicating answer correctness, such measures are heavily dataset-dependent and are unlikely to generalize. In this work, we begin by investigating the hidden representations of questions, answers, and contexts in transformer-based QA architectures. We observe a consistent pattern in the answer representations, which we show can be used to automatically evaluate whether or not a predicted answer span is correct. Our method does not require any labeled data and outperforms strong heuristic baselines, across 2 datasets and 7 domains. We are able to predict whether or not a model's answer is correct with 91.37% accuracy on SQuAD, and 80.7% accuracy on SubjQA. We expect that this method will have broad applications, e.g., in the semi-automatic development of QA datasets