Text articles with false claims, especially news, have recently become aggravating for the Internet users. These articles are in wide circulation and readers face difficulty discerning fact from fiction. Previous work on credibility assessment has focused on factual analysis and linguistic features. The task's main challenge is the distinction between the features of true and false articles. In this paper, we propose a novel approach called Credibility Outcome (CREDO) which aims at scoring the credibility of an article in an open domain setting. CREDO consists of different modules for capturing various features responsible for the credibility of an article. These features includes credibility of the article's source and author, semantic similarity between the article and related credible articles retrieved from a knowledge base, and sentiments conveyed by the article. A neural network architecture learns the contribution of each of these modules to the overall credibility of an article. Experiments on Snopes dataset reveals that CREDO outperforms the state-of-the-art approaches based on linguistic features.

The advent of data-driven real-time applications requires the implementation of Deep Neural Networks (DNNs) on Machine Learning accelerators. Google's Tensor Processing Unit (TPU) is one such neural network accelerator that uses systolic array-based matrix multiplication hardware for computation in its crux. Manufacturing faults at any state element of the matrix multiplication unit can cause unexpected errors in these inference networks. In this paper, we propose a formal model of permanent faults and their propagation in a TPU using the Discrete-Time Markov Chain (DTMC) formalism. The proposed model is analyzed using the probabilistic model checking technique to reason about the likelihood of faulty outputs. The obtained quantitative results show that the classification accuracy is sensitive to the type of permanent faults as well as their location, bit position and the number of layers in the neural network. The conclusions from our theoretical model have been validated using experiments on a digit recognition-based DNN.

To demystify the "black box" property of deep neural networks for natural language processing (NLP), several methods have been proposed to interpret their predictions by measuring the change in prediction probability after erasing each token of an input. Since existing methods replace each token with a predefined value (i.e., zero), the resulting sentence lies out of the training data distribution, yielding misleading interpretations. In this study, we raise the out-of-distribution problem induced by the existing interpretation methods and present a remedy; we propose to marginalize each token out. We interpret various NLP models trained for sentiment analysis and natural language inference using the proposed method.

Detecting if a text is humorous is a hard task to do computationally, as it usually requires linguistic and common sense insights. In machine learning, humor detection is usually modeled as a binary classification task, trained to predict if the given text is a joke or another type of text. Rather than using completely different non-humorous texts, we propose using text generation algorithms for imitating the original joke dataset to increase the difficulty for the learning algorithm. We constructed several different joke and non-joke datasets to test the humor detection abilities of different language technologies. In particular, we compare the humor detection capabilities of classic neural network approaches with the state-of-the-art Dutch language model RobBERT. In doing so, we create and compare the first Dutch humor detection systems. We found that while other language models perform well when the non-jokes came from completely different domains, RobBERT was the only one that was able to distinguish jokes from generated negative examples. This performance illustrates the usefulness of using text generation to create negative datasets for humor recognition, and also shows that transformer models are a large step forward in humor detection.

Differentiation lies at the core of many machine-learning algorithms, and is well-supported by popular autodiff systems, such as TensorFlow and PyTorch. Originally, these systems have been developed to compute derivatives of differentiable functions, but in practice, they are commonly applied to functions with non-differentiabilities. For instance, neural networks using ReLU define non-differentiable functions in general, but the gradients of losses involving those functions are computed using autodiff systems in practice. This status quo raises a natural question: are autodiff systems correct in any formal sense when they are applied to such non-differentiable functions? In this paper, we provide a positive answer to this question. Using counterexamples, we first point out flaws in often-used informal arguments, such as: non-differentiabilities arising in deep learning do not cause any issues because they form a measure-zero set. We then investigate a class of functions, called PAP functions, that includes nearly all (possibly non-differentiable) functions in deep learning nowadays. For these PAP functions, we propose a new type of derivatives, called intensional derivatives, and prove that these derivatives always exist and coincide with standard derivatives for almost all inputs. We also show that these intensional derivatives are what most autodiff systems compute or try to compute essentially. In this way, we formally establish the correctness of autodiff systems applied to non-differentiable functions.

State-of-the-art optimization is steadily shifting towards massively parallel pipelines with extremely large batch sizes. As a consequence, CPU-bound preprocessing and disk/memory/network operations have emerged as new performance bottlenecks, as opposed to hardware-accelerated gradient computations. In this regime, a recently proposed approach is data echoing (Choi et al., 2019), which takes repeated gradient steps on the same batch while waiting for fresh data to arrive from upstream. We provide the first convergence analyses of "data-echoed" extensions of common optimization methods, showing that they exhibit provable improvements over their synchronous counterparts. Specifically, we show that in convex optimization with stochastic minibatches, data echoing affords speedups on the curvature-dominated part of the convergence rate, while maintaining the optimal statistical rate.

We outline the design and construction of novel robotic arms using machine perception, convolutional neural networks, and symbolic AI for logical control and affordance indexing. We describe our robotic arms built with a humanlike mechanical configuration and aesthetic, with 28 degrees of freedom, touch sensors, and series elastic actuators. The arms were modelled in Roodle and Gazebo with URDF models, as well as Unity, and implement motion control solutions for solving live games of Baccarat (the casino card game), rock paper scissors, handshaking, and drawing. This includes live interactions with people, incorporating both social control of the hands and facial gestures, and physical inverse kinematics (IK) for grasping and manipulation tasks. The resulting framework is an integral part of the Sophia 2020 alpha platform, which is being used with ongoing research in the authors work with team AHAM, an ANA Avatar Xprize effort towards human-AI hybrid telepresence. These results are available to test on the broadly released Hanson Robotics Sophia 2020 robot platform, for users to try and extend.

Ontology Alignment is an important research problem that finds application in various fields such as data integration, data transfer, data preparation etc. State-of-the-art (SOTA) architectures in Ontology Alignment typically use naive domain-dependent approaches with handcrafted rules and manually assigned values, making them unscalable and inefficient. Deep Learning approaches for ontology alignment use domain-specific architectures that are not only in-extensible to other datasets and domains, but also typically perform worse than rule-based approaches due to various limitations including over-fitting of models, sparsity of datasets etc. In this work, we propose VeeAlign, a Deep Learning based model that uses a dual-attention mechanism to compute the contextualized representation of a concept in order to learn alignments. By doing so, not only does our approach exploit both syntactic and semantic structure of ontologies, it is also, by design, flexible and scalable to different domains with minimal effort. We validate our approach on various datasets from different domains and in multilingual settings, and show its superior performance over SOTA methods.

Objective dyslexia diagnosis is not a straighforward task since it is traditionally performed by means of the intepretation of different behavioural tests. Moreover, these tests are only applicable to readers. This way, early diagnosis requires the use of specific tasks not only related to reading. Thus, the use of Electroencephalography (EEG) constitutes an alternative for an objective and early diagnosis that can be used with pre-readers. In this way, the extraction of relevant features in EEG signals results crucial for classification. However, the identification of the most relevant features is not straighforward, and predefined statistics in the time or frequency domain are not always discriminant enough. On the other hand, classical processing of EEG signals based on extracting EEG bands frequency descriptors, usually make some assumptions on the raw signals that could cause indormation loosing. In this work we propose an alternative for analysis in the frequency domain based on Singluar Spectrum Analysis (SSA) to split the raw signal into components representing different oscillatory modes. Moreover, correlation matrices obtained for each component among EEG channels are classfied using a Convolutional Neural network.

Transfer learning is a useful machine learning framework that allows one to build task-specific models (student models) without significantly incurring training costs using a single powerful model (teacher model) pre-trained with a large amount of data. The teacher model may contain private data, or interact with private inputs. We investigate if one can leak or infer such private information without interacting with the teacher model directly. We describe such inference attacks in the context of face recognition, an application of transfer learning that is highly sensitive to personal privacy. Under black-box and realistic settings, we show that existing inference techniques are ineffective, as interacting with individual training instances through the student models does not reveal information about the teacher. We then propose novel strategies to infer from aggregate-level information. Consequently, membership inference attacks on the teacher model are shown to be possible, even when the adversary has access only to the student models. We further demonstrate that sensitive attributes can be inferred, even in the case where the adversary has limited auxiliary information. Finally, defensive strategies are discussed and evaluated. Our extensive study indicates that information leakage is a real privacy threat to the transfer learning framework widely used in real-life situations.