This paper demonstrates a fatal vulnerability in natural language inference (NLI) and text classification systems. More concretely, we present a 'backdoor poisoning' attack on NLP models. Our poisoning attack utilizes conditional adversarially regularized autoencoder (CARA) to generate poisoned training samples by poison injection in latent space. Just by adding 1% poisoned data, our experiments show that a victim BERT finetuned classifier's predictions can be steered to the poison target class with success rates of >80% when the input hypothesis is injected with the poison signature, demonstrating that NLI and text classification systems face a huge security risk.

There have been significant efforts to interpret the encoder of Transformer-based encoder-decoder architectures for neural machine translation (NMT); meanwhile, the decoder remains largely unexamined despite its critical role. During translation, the decoder must predict output tokens by considering both the source-language text from the encoder and the target-language prefix produced in previous steps. In this work, we study how Transformer-based decoders leverage information from the source and target languages -- developing a universal probe task to assess how information is propagated through each module of each decoder layer. We perform extensive experiments on three major translation datasets (WMT En-De, En-Fr, and En-Zh). Our analysis provides insight on when and where decoders leverage different sources. Based on these insights, we demonstrate that the residual feed-forward module in each Transformer decoder layer can be dropped with minimal loss of performance -- a significant reduction in computation and number of parameters, and consequently a significant boost to both training and inference speed.

We propose a method for improving adversarial robustness by addition of a new bounded function just before softmax. Recent studies hypothesize that small logits (inputs of softmax) by logit regularization can improve adversarial robustness of deep learning. Following this hypothesis, we analyze norms of logit vectors at the optimal point under the assumption of universal approximation and explore new methods for constraining logits by addition of a bounded function before softmax. We theoretically and empirically reveal that small logits by addition of a common activation function, e.g., hyperbolic tangent, do not improve adversarial robustness since input vectors of the function (pre-logit vectors) can have large norms. From the theoretical findings, we develop the new bounded function. The addition of our function improves adversarial robustness because it makes logit and pre-logit vectors have small norms. Since our method only adds one activation function before softmax, it is easy to combine our method with adversarial training. Our experiments demonstrate that our method is comparable to logit regularization methods in terms of accuracies on adversarially perturbed datasets without adversarial training. Furthermore, it is superior or comparable to logit regularization methods and a recent defense method (TRADES) when using adversarial training.

Flow-based generative models have become an important class of unsupervised learning approaches. In this work, we incorporate the key idea of renormalization group (RG) and sparse prior distribution to design a hierarchical flow-based generative model, called RG-Flow, which can separate different scale information of images with disentangle representations at each scale. We demonstrate our method mainly on the CelebA dataset and show that the disentangled representation at different scales enables semantic manipulation and style mixing of the images. To visualize the latent representation, we introduce the receptive fields for flow-based models and find receptive fields learned by RG-Flow are similar to convolutional neural networks. In addition, we replace the widely adopted Gaussian prior distribution by sparse prior distributions to further enhance the disentanglement of representations. One of the most important unsupervised learning tasks is to learn the data distribution and build generative models. Over the past few years, various types of generative models have been proposed. Yet the latent variables are on equal footing and mixed globally. Here, we propose a new flow-based model, RG-Flow, which is inspired by the idea of renormalization group in statistical physics. RG-Flow imposes locality and hierarchical structure in bijective transformations. It allows us to access different scale information in original images by latent variables at different locations, which offers better explainability.

Subjectivity is the expression of internal opinions or beliefs which cannot be objectively observed or verified, and has been shown to be important for sentiment analysis and word-sense disambiguation. Furthermore, subjectivity is an important aspect of user-generated data. In spite of this, subjectivity has not been investigated in contexts where such data is widespread, such as in question answering (QA). We therefore investigate the relationship between subjectivity and QA, while developing a new dataset. We compare and contrast with analyses from previous work, and verify that findings regarding subjectivity still hold when using recently developed NLP architectures. We find that subjectivity is also an important feature in the case of QA, albeit with more intricate interactions between subjectivity and QA performance. For instance, a subjective question may or may not be associated with a subjective answer. We release an English QA dataset (SubjQA) based on customer reviews, containing subjectivity annotations for questions and answer spans across 6 distinct domains.

Visual dialog is a challenging vision-language task in which a series of questions visually grounded by a given image are answered. To resolve the visual dialog task, a high-level understanding of various multimodal inputs (e.g., question, dialog history, and image) is required. Specifically, it is necessary for an agent to 1) determine the semantic intent of question and 2) align question-relevant textual and visual contents among heterogeneous modality inputs. In this paper, we propose Multi-View Attention Network (MVAN), which leverages multiple views about heterogeneous inputs based on attention mechanisms. MVAN effectively captures the question-relevant information from the dialog history with two complementary modules (i.e., Topic Aggregation and Context Matching), and builds multimodal representations through sequential alignment processes (i.e., Modality Alignment). Experimental results on VisDial v1.0 dataset show the effectiveness of our proposed model, which outperforms the previous state-of-the-art methods with respect to all evaluation metrics.

AWS DeepComposer provides a creative and hands-on experience for learning generative AI and machine learning (ML). We recently launched the Edit melody feature, which allows you to add, remove, or edit specific notes, giving you full control of the pitch, length, and timing for each note. In this post, you can learn to use the Edit melody feature to collaborate with the autoregressive convolutional neural network (AR-CNN) algorithm and create interesting Bach-style compositions. Through human-AI collaboration, we can surpass what humans and AI systems can create independently. For example, you can seek inspiration from AI to create art or music outside their area of expertise or offload the more routine tasks, like creating variations on a melody, and focus on the more interesting and creative tasks.

Most of the recently trending technologies such as BERT, GPT-3, Transformers, LSTM, GANs and others have deep learning at the core. These deep learning-based applications are transforming many industries such as self-driving, language translation, fraud detection and more. The researchers in the field of deep learning are contributing immensely to bring some fantastic applications in the field. In this article, we list ten deep learning researchers, in no particular order, who are re-defining the application areas of deep learning. A pioneer in deep learning and machine learning-based research, Hinton's work is aimed at finding complex structure in large, high-dimensional datasets, and understanding how the brain learns to see.

Data is rampant in the Internet of Things (IoT) Age, during which the exponential growth of data has outpaced the capacity of traditional computing. It has reached maturity in some stages, but it is in adolescence in new stages. With the advent of Big Data, businesses are merging it with big compute and IoT for analytics using Artificial Intelligence (AI). After the raw input from the big data gets cleaned, structured, and unified, AI performs cognitive functions and outputs values for the business. With its ability to analyze massive amounts of data in milliseconds, it can now be processed in "real-time".

Machine Learning is an area of research that allows machines the ability to learn without being directly programmed. Machine Learning development is in trend as many students, teachers, developers, and data scientists use machine learning to develop various projects and products. However, developing machine learning models require high system requirement specifications as sometimes the model training process can go from 2 hours to 2 days and more. So low-end systems can not handle training of good machine learning models or even if they somehow train models, critical system issues are likely to occur. However, there are many open-source Machine Learning environments available that do not require any system requirement specification and use cloud infrastructure to train your model in the most optimal time possible.