This paper develops three methods for visually perturbing text to attack NLP systems, confirms that humans are robust to such perturbations but NLP performance degrades, and proposes three methods for shielding systems from visual attacks. Specifically:
- It introduces three visual perturbation methods: image-based, description-based, and easy-character replacements.
- Human tests show humans have low error rates understanding perturbed text, while NLP tasks like POS tagging and toxicity classification suffer reduced accuracy.
- To shield systems, it employs adversarial training, initializing models with visual embeddings, and rule-based character recovery - which improve performance over unshielded models under attack.
This document discusses deep learning applications for natural language processing (NLP). It begins by explaining what deep learning and deep neural networks are, and how they build upon older neural network models by adding multiple hidden layers. It then discusses why deep learning is now more viable due to factors like increased computational power from GPUs and improved training methods. The document outlines several NLP tasks that benefit from deep learning techniques, such as word embeddings, dependency parsing, sentiment analysis. It also provides examples of tools used for deep learning NLP and discusses building a sentence classifier to identify funding sentences from news articles.
The document discusses word embedding techniques, specifically Word2vec. It introduces the motivation for distributed word representations and describes the Skip-gram and CBOW architectures. Word2vec produces word vectors that encode linguistic regularities, with simple examples showing words with similar relationships have similar vector offsets. Evaluation shows Word2vec outperforms previous methods, and its word vectors are now widely used in NLP applications.
This document provides an overview of deep learning techniques for natural language processing. It begins with an introduction to distributed word representations like word2vec and GloVe. It then discusses methods for generating sentence embeddings, including paragraph vectors and recursive neural networks. Character-level models are presented as an alternative to word embeddings that can handle morphology and out-of-vocabulary words. Finally, some general deep learning approaches for NLP tasks like text generation and word sense disambiguation are briefly outlined.
This document provides an outline for a tutorial on deep learning for natural language processing. It begins with an introduction to deep learning and its history, then discusses how neural methods have become prominent in natural language processing. The rest of the tutorial is outlined covering deep semantic models for text, recurrent neural networks for text generation, neural question answering models, and deep reinforcement learning for dialog systems.
Convolutional neural networks (CNNs) have traditionally been used for computer vision tasks but recent work has applied them to language modeling as well. CNNs treat sequences of words as signals over time rather than independent units. They use convolution and pooling layers to identify important n-gram features. Results show CNNs can be effective for classification tasks like sentiment analysis but have had less success with sequence modeling tasks. Overall, CNNs provide an alternative to recurrent neural networks for certain natural language processing problems and help understand each model's strengths and weaknesses.
StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery ivaderivader
The paper presents three methods for text-driven manipulation of StyleGAN imagery using CLIP:
1. Direct optimization of the latent w vector to match a text prompt
2. Training a mapping function to map text to changes in the latent space
3. Finding global directions in the latent space corresponding to attributes by measuring distances between text embeddings
The methods allow editing StyleGAN images based on natural language instructions and demonstrate CLIP's ability to provide fine-grained controls, but rely on pretrained StyleGAN and CLIP models and may struggle with unseen text or image domains.
The document discusses image captioning using deep neural networks. It begins by providing examples of how humans can easily describe images but generating image captions with a computer program was previously very difficult. Recent advances in deep learning, specifically using convolutional neural networks (CNNs) to recognize objects in images and recurrent neural networks (RNNs) to generate captions, have enabled automated image captioning. The document discusses CNN and RNN architectures for image captioning and provides examples of pre-trained models that can be used, such as VGG-16.
Contrastive Learning with Adversarial Perturbations for Conditional Text Gene...MLAI2
Recently, sequence-to-sequence (seq2seq) models with the Transformer architecture have achieved remarkable performance on various conditional text generation tasks, such as machine translation. However, most of them are trained with teacher forcing with the ground truth label given at each time step, without being exposed to incorrectly generated tokens during training, which hurts its generalization to unseen inputs, that is known as the “exposure bias” problem. In this work, we propose to mitigate the conditional text generation problem by contrasting positive pairs with negative pairs, such that the model is exposed to various valid or incorrect perturbations of the inputs, for improved generalization. However, training the model with naïve contrastive learning framework using random non-target sequences as negative examples is suboptimal, since they are easily distinguishable from the correct output, especially so with models pretrained with large text corpora. Also, generating positive examples requires domain-specific augmentation heuristics which may not generalize over diverse domains. To tackle this problem, we propose a principled method to generate positive and negative samples for contrastive learning of seq2seq models. Specifically, we generate negative examples by adding small perturbations to the input sequence to minimize its conditional likelihood, and positive examples by adding large perturbations while enforcing it to have a high conditional likelihood. Such “hard” positive and negative pairs generated using our method guides the model to better distinguish correct outputs from incorrect ones. We empirically show that our proposed method significantly improves the generalization of the seq2seq on three text generation tasks — machine translation, text summarization, and question generation.
This document discusses deep learning applications for natural language processing (NLP). It begins by explaining what deep learning and deep neural networks are, and how they build upon older neural network models by adding multiple hidden layers. It then discusses why deep learning is now more viable due to factors like increased computational power from GPUs and improved training methods. The document outlines several NLP tasks that benefit from deep learning techniques, such as word embeddings, dependency parsing, sentiment analysis. It also provides examples of tools used for deep learning NLP and discusses building a sentence classifier to identify funding sentences from news articles.
The document discusses word embedding techniques, specifically Word2vec. It introduces the motivation for distributed word representations and describes the Skip-gram and CBOW architectures. Word2vec produces word vectors that encode linguistic regularities, with simple examples showing words with similar relationships have similar vector offsets. Evaluation shows Word2vec outperforms previous methods, and its word vectors are now widely used in NLP applications.
This document provides an overview of deep learning techniques for natural language processing. It begins with an introduction to distributed word representations like word2vec and GloVe. It then discusses methods for generating sentence embeddings, including paragraph vectors and recursive neural networks. Character-level models are presented as an alternative to word embeddings that can handle morphology and out-of-vocabulary words. Finally, some general deep learning approaches for NLP tasks like text generation and word sense disambiguation are briefly outlined.
This document provides an outline for a tutorial on deep learning for natural language processing. It begins with an introduction to deep learning and its history, then discusses how neural methods have become prominent in natural language processing. The rest of the tutorial is outlined covering deep semantic models for text, recurrent neural networks for text generation, neural question answering models, and deep reinforcement learning for dialog systems.
Convolutional neural networks (CNNs) have traditionally been used for computer vision tasks but recent work has applied them to language modeling as well. CNNs treat sequences of words as signals over time rather than independent units. They use convolution and pooling layers to identify important n-gram features. Results show CNNs can be effective for classification tasks like sentiment analysis but have had less success with sequence modeling tasks. Overall, CNNs provide an alternative to recurrent neural networks for certain natural language processing problems and help understand each model's strengths and weaknesses.
StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery ivaderivader
The paper presents three methods for text-driven manipulation of StyleGAN imagery using CLIP:
1. Direct optimization of the latent w vector to match a text prompt
2. Training a mapping function to map text to changes in the latent space
3. Finding global directions in the latent space corresponding to attributes by measuring distances between text embeddings
The methods allow editing StyleGAN images based on natural language instructions and demonstrate CLIP's ability to provide fine-grained controls, but rely on pretrained StyleGAN and CLIP models and may struggle with unseen text or image domains.
The document discusses image captioning using deep neural networks. It begins by providing examples of how humans can easily describe images but generating image captions with a computer program was previously very difficult. Recent advances in deep learning, specifically using convolutional neural networks (CNNs) to recognize objects in images and recurrent neural networks (RNNs) to generate captions, have enabled automated image captioning. The document discusses CNN and RNN architectures for image captioning and provides examples of pre-trained models that can be used, such as VGG-16.
Contrastive Learning with Adversarial Perturbations for Conditional Text Gene...MLAI2
Recently, sequence-to-sequence (seq2seq) models with the Transformer architecture have achieved remarkable performance on various conditional text generation tasks, such as machine translation. However, most of them are trained with teacher forcing with the ground truth label given at each time step, without being exposed to incorrectly generated tokens during training, which hurts its generalization to unseen inputs, that is known as the “exposure bias” problem. In this work, we propose to mitigate the conditional text generation problem by contrasting positive pairs with negative pairs, such that the model is exposed to various valid or incorrect perturbations of the inputs, for improved generalization. However, training the model with naïve contrastive learning framework using random non-target sequences as negative examples is suboptimal, since they are easily distinguishable from the correct output, especially so with models pretrained with large text corpora. Also, generating positive examples requires domain-specific augmentation heuristics which may not generalize over diverse domains. To tackle this problem, we propose a principled method to generate positive and negative samples for contrastive learning of seq2seq models. Specifically, we generate negative examples by adding small perturbations to the input sequence to minimize its conditional likelihood, and positive examples by adding large perturbations while enforcing it to have a high conditional likelihood. Such “hard” positive and negative pairs generated using our method guides the model to better distinguish correct outputs from incorrect ones. We empirically show that our proposed method significantly improves the generalization of the seq2seq on three text generation tasks — machine translation, text summarization, and question generation.
Colloquium talk on modal sense classification using a convolutional neural ne...Ana Marasović
Modal sense classification (MSC) is a special case of sense disambiguation relevant for distinguishing facts from hypotheses and speculations, or apprehended, planned and desired states of affairs. Prior approaches showed that even with carefully designed semantic feature sets, the models have difficulties beating the majority sense baseline in cases of difficult sense distinctions and when applying the models to heterogeneous text genres. Another drawback of former approaches is that feature implementation heavily depends on a external language-specific resources such as dependency or constituency parse trees and lexical databases such as WordNet or CELEX. To alleviate manual crafting of the features and to obtain a model which is easily portable to novel languages, we propose to cast MSC as a sentence classification task with a fixed sense inventory in a convolutional neural network (CNN) architecture. Our performance study shows that CNN is an appropriate model for MSC and its special properties motivate us to investigate it as a formal framework for general word sense disambiguation tasks.
Deep Learning & NLP: Graphs to the Rescue!Roelof Pieters
This document provides an overview of deep learning and natural language processing techniques. It begins with a history of machine learning and how deep learning advanced beyond early neural networks using methods like backpropagation. Deep learning methods like convolutional neural networks and word embeddings are discussed in the context of natural language processing tasks. Finally, the document proposes some graph-based approaches to combining deep learning with NLP, such as encoding language structures in graphs or using finite state graphs trained with genetic algorithms.
- High-level overview
- Challenges in natural language processing
- What is intelligence?
- Sequence prediction
- A very short history of Solomonoff induction
- Meaning acquisition
- Logistic loss
- Gradient descent
- Applications
It's a brief overview of Natural Language Processing using Python module NLTK.The codes for demonstration can be found from the github link given in the references slide.
This document summarizes a paper on using simple lexical overlap features with support vector machines (SVMs) for Russian paraphrase identification. It introduces paraphrase identification and various paraphrase corpora. It then describes a knowledge-lean approach using only tokenization, lowercasing, and overlap features like union and intersection size as inputs to linear and RBF kernel SVMs. The method achieves competitive results on English, Turkish, and Russian paraphrase identification tasks.
[Paper Reading] Unsupervised Learning of Sentence Embeddings using Compositi...Hiroki Shimanaka
(1) The document presents an unsupervised method called Sent2Vec to learn sentence embeddings using compositional n-gram features. (2) Sent2Vec extends the continuous bag-of-words model to train sentence embeddings by composing word vectors with n-gram embeddings. (3) Experimental results show Sent2Vec outperforms other unsupervised models on most benchmark tasks, highlighting the robustness of the sentence embeddings produced.
This document discusses natural language processing and language models. It begins by explaining that natural language processing aims to give computers the ability to process human language in order to perform tasks like dialogue systems, machine translation, and question answering. It then discusses how language models assign probabilities to strings of text to determine if they are valid sentences. Specifically, it covers n-gram models which use the previous n words to predict the next, and how smoothing techniques are used to handle uncommon words. The document provides an overview of key concepts in natural language processing and language modeling.
Deep Learning for Information Retrieval: Models, Progress, & OpportunitiesMatthew Lease
Talk given at the 8th Forum for Information Retrieval Evaluation (FIRE, http://fire.irsi.res.in/fire/2016/), December 10, 2016, and at the Qatar Computing Research Institute (QCRI), December 15, 2016.
Recent studies on robustness of Convolutional Neural Networks (CNN) shows that CNNs are highly vulnerable towards adversarial attacks. Meanwhile, smaller sized CNN models with no signicant accuracy loss are being introduced to mobile devices. However, only the accuracy on standard datasets is reported along with such research. The wide deployment of smaller models on millions of mobile devices stresses importance of their robustness. In this research, we study how robust such models are with respect to state-of-the-art compression techniques such as quantization.
Deep generative models can generate synthetic images, speech, text and other data types. There are three popular types: autoregressive models which generate data step-by-step; variational autoencoders which learn the distribution of latent variables to generate data; and generative adversarial networks which train a generator and discriminator in an adversarial game to generate high quality samples. Generative models have applications in image generation, translation between domains, and simulation.
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 Saurabh Kaushik
This document discusses how deep learning techniques can be applied to natural language processing tasks. It begins by explaining some of the limitations of traditional rule-based and machine learning approaches to NLP, such as the lack of semantic understanding and difficulty of feature engineering. Deep learning approaches can learn features automatically from large amounts of unlabeled text and better capture semantic and syntactic relationships between words. Recurrent neural networks are well-suited for NLP because they can model sequential data like text, and convolutional neural networks can learn hierarchical patterns in text.
BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingSeonghyun Kim
The document discusses BERT, which stands for Bidirectional Encoder Representations from Transformers. BERT uses bidirectional Transformers to pre-train deep contextual representations of language. It was trained on two unsupervised prediction tasks using large text corpora. Experimental results showed that BERT achieved state-of-the-art results on eleven natural language understanding tasks, including question answering and textual inference. The document outlines the model architecture of BERT and the pre-training and fine-tuning methods used.
Enhancing Entity Linking by Combining NER ModelsJulien PLU
The document describes enhancements made to the ADEL entity linking framework. ADEL combines multiple named entity recognition models and uses a combination of linguistic and dictionary-based approaches. New features in ADEL include using a generic API to interface with NLP tools, combining multiple CRF models for entity extraction, clustering nil entities, and developing a new backend using Elasticsearch and Couchbase. The document compares the performance of the original 2015 version of ADEL to the new 2016 version on standard entity linking tasks and datasets.
In this talk, you will discover how the 15k LOC codebase was implemented with spec so you don't have to (but probably should). Validation; testing; destructuring; composable “data macros” via conformers; we’ve tried spec in all its multifaceted glory. You will discover a distillation of lessons learned interspersed with musing on how spec alters development flow and one’s thinking.
Machine learning on graphs is an important and ubiquitous task with applications ranging from drug design to friendship recommendation in social networks. The primary challenge in this domain is finding a way to represent, or encode, graph structure so that it can be easily exploited by machine learning models. However, traditionally machine learning approaches relied on user-defined heuristics to extract features encoding structural information about a graph. In this talk I will discuss methods that automatically learn to encode graph structure into low-dimensional embeddings, using techniques based on deep learning and nonlinear dimensionality reduction. I will provide a conceptual review of key advancements in this area of representation learning on graphs, including random-walk based algorithms, and graph convolutional networks.
KOSMOS-1 is a multimodal large language model that can perceive and process language as well as visual inputs like images. It was trained on large web-scale datasets containing text, images, and image-caption pairs to align its vision capabilities with its natural language understanding. Experimental results showed that KOSMOS-1 can perform well on tasks involving language, vision, and their combination, including image captioning, visual question answering, and describing images based on text instructions, all without any fine-tuning. The ability to perceive and understand different modalities allows language models to acquire knowledge in new ways and expands their application to areas like robotics and document intelligence.
[Mmlab seminar 2016] deep learning for human pose estimationWei Yang
This document summarizes recent advances in deep learning approaches for human pose estimation. It describes early methods like DeepPose that used cascades of regressors. Later works introduced heatmap regression to capture spatial information. Convolutional Pose Machine and Stacked Hourglass networks further improved accuracy by incorporating stronger context modeling through deeper networks with larger receptive fields and intermediate supervision. These approaches demonstrate that both local appearance cues and modeling of global context and structure are important for accurate human pose estimation.
What Does the Webinar Cover?
You'll learn how to optimize varying parameters and disciplines throughout the lifecycle of the system within cost and schedule constraints without compromising performance. Real MBSE enables the execution of many activities in parallel, thus enabling the “faster and cheaper” part.
Many people can contribute to the design and development at the same time, because the information they create can be easily linked together to form abstractions that enable you to communicate the results at all levels. Dr. Dam uses a methodology that includes the technique, processes, and tools.
This methodology isn’t the only way to have a successful MBSE capability, but all three elements must be incorporated in any methodology you use. We offer this methodology as one that has proven successful over the past decade. It is based on methodologies used since the 1960s, but updated to the modern cloud computing, artificial intelligence age; that's now emerging toward the end of the second decade of the 21st Century.
Often people today work in a similar manner to how their grandparents worked in the 1960s, just with electronic tools instead of paper and pencil. Just creating a “model” doesn’t mean you are doing effective MBSE. This webinar will show you how to take MBSE into the 21st century.
Natural Language Understanding of Systems Engineering ArtifactsÁkos Horváth
This paper examines in close relation two fields of growing importance: model-based systems engineering (MBSE) and natural language processing (NLP). System models provide a structured description of engineering data, whose inherent semantics often remains hard to explore. Natural language understanding, (i.e., the machine analysis of texts produced by humans) an important field of NLP, focuses on semantic text comprehension but cannot directly account for structured information sources.
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2024/06/temporal-event-neural-networks-a-more-efficient-alternative-to-the-transformer-a-presentation-from-brainchip/
Chris Jones, Director of Product Management at BrainChip , presents the “Temporal Event Neural Networks: A More Efficient Alternative to the Transformer” tutorial at the May 2024 Embedded Vision Summit.
The expansion of AI services necessitates enhanced computational capabilities on edge devices. Temporal Event Neural Networks (TENNs), developed by BrainChip, represent a novel and highly efficient state-space network. TENNs demonstrate exceptional proficiency in handling multi-dimensional streaming data, facilitating advancements in object detection, action recognition, speech enhancement and language model/sequence generation. Through the utilization of polynomial-based continuous convolutions, TENNs streamline models, expedite training processes and significantly diminish memory requirements, achieving notable reductions of up to 50x in parameters and 5,000x in energy consumption compared to prevailing methodologies like transformers.
Integration with BrainChip’s Akida neuromorphic hardware IP further enhances TENNs’ capabilities, enabling the realization of highly capable, portable and passively cooled edge devices. This presentation delves into the technical innovations underlying TENNs, presents real-world benchmarks, and elucidates how this cutting-edge approach is positioned to revolutionize edge AI across diverse applications.
Have you ever been confused by the myriad of choices offered by AWS for hosting a website or an API?
Lambda, Elastic Beanstalk, Lightsail, Amplify, S3 (and more!) can each host websites + APIs. But which one should we choose?
Which one is cheapest? Which one is fastest? Which one will scale to meet our needs?
Join me in this session as we dive into each AWS hosting service to determine which one is best for your scenario and explain why!
More Related Content
Similar to Text Processing Like Humans Do: Visually Attacking and Shielding NLP Systems
Colloquium talk on modal sense classification using a convolutional neural ne...Ana Marasović
Modal sense classification (MSC) is a special case of sense disambiguation relevant for distinguishing facts from hypotheses and speculations, or apprehended, planned and desired states of affairs. Prior approaches showed that even with carefully designed semantic feature sets, the models have difficulties beating the majority sense baseline in cases of difficult sense distinctions and when applying the models to heterogeneous text genres. Another drawback of former approaches is that feature implementation heavily depends on a external language-specific resources such as dependency or constituency parse trees and lexical databases such as WordNet or CELEX. To alleviate manual crafting of the features and to obtain a model which is easily portable to novel languages, we propose to cast MSC as a sentence classification task with a fixed sense inventory in a convolutional neural network (CNN) architecture. Our performance study shows that CNN is an appropriate model for MSC and its special properties motivate us to investigate it as a formal framework for general word sense disambiguation tasks.
Deep Learning & NLP: Graphs to the Rescue!Roelof Pieters
This document provides an overview of deep learning and natural language processing techniques. It begins with a history of machine learning and how deep learning advanced beyond early neural networks using methods like backpropagation. Deep learning methods like convolutional neural networks and word embeddings are discussed in the context of natural language processing tasks. Finally, the document proposes some graph-based approaches to combining deep learning with NLP, such as encoding language structures in graphs or using finite state graphs trained with genetic algorithms.
- High-level overview
- Challenges in natural language processing
- What is intelligence?
- Sequence prediction
- A very short history of Solomonoff induction
- Meaning acquisition
- Logistic loss
- Gradient descent
- Applications
It's a brief overview of Natural Language Processing using Python module NLTK.The codes for demonstration can be found from the github link given in the references slide.
This document summarizes a paper on using simple lexical overlap features with support vector machines (SVMs) for Russian paraphrase identification. It introduces paraphrase identification and various paraphrase corpora. It then describes a knowledge-lean approach using only tokenization, lowercasing, and overlap features like union and intersection size as inputs to linear and RBF kernel SVMs. The method achieves competitive results on English, Turkish, and Russian paraphrase identification tasks.
[Paper Reading] Unsupervised Learning of Sentence Embeddings using Compositi...Hiroki Shimanaka
(1) The document presents an unsupervised method called Sent2Vec to learn sentence embeddings using compositional n-gram features. (2) Sent2Vec extends the continuous bag-of-words model to train sentence embeddings by composing word vectors with n-gram embeddings. (3) Experimental results show Sent2Vec outperforms other unsupervised models on most benchmark tasks, highlighting the robustness of the sentence embeddings produced.
This document discusses natural language processing and language models. It begins by explaining that natural language processing aims to give computers the ability to process human language in order to perform tasks like dialogue systems, machine translation, and question answering. It then discusses how language models assign probabilities to strings of text to determine if they are valid sentences. Specifically, it covers n-gram models which use the previous n words to predict the next, and how smoothing techniques are used to handle uncommon words. The document provides an overview of key concepts in natural language processing and language modeling.
Deep Learning for Information Retrieval: Models, Progress, & OpportunitiesMatthew Lease
Talk given at the 8th Forum for Information Retrieval Evaluation (FIRE, http://fire.irsi.res.in/fire/2016/), December 10, 2016, and at the Qatar Computing Research Institute (QCRI), December 15, 2016.
Recent studies on robustness of Convolutional Neural Networks (CNN) shows that CNNs are highly vulnerable towards adversarial attacks. Meanwhile, smaller sized CNN models with no signicant accuracy loss are being introduced to mobile devices. However, only the accuracy on standard datasets is reported along with such research. The wide deployment of smaller models on millions of mobile devices stresses importance of their robustness. In this research, we study how robust such models are with respect to state-of-the-art compression techniques such as quantization.
Deep generative models can generate synthetic images, speech, text and other data types. There are three popular types: autoregressive models which generate data step-by-step; variational autoencoders which learn the distribution of latent variables to generate data; and generative adversarial networks which train a generator and discriminator in an adversarial game to generate high quality samples. Generative models have applications in image generation, translation between domains, and simulation.
Engineering Intelligent NLP Applications Using Deep Learning – Part 2 Saurabh Kaushik
This document discusses how deep learning techniques can be applied to natural language processing tasks. It begins by explaining some of the limitations of traditional rule-based and machine learning approaches to NLP, such as the lack of semantic understanding and difficulty of feature engineering. Deep learning approaches can learn features automatically from large amounts of unlabeled text and better capture semantic and syntactic relationships between words. Recurrent neural networks are well-suited for NLP because they can model sequential data like text, and convolutional neural networks can learn hierarchical patterns in text.
BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingSeonghyun Kim
The document discusses BERT, which stands for Bidirectional Encoder Representations from Transformers. BERT uses bidirectional Transformers to pre-train deep contextual representations of language. It was trained on two unsupervised prediction tasks using large text corpora. Experimental results showed that BERT achieved state-of-the-art results on eleven natural language understanding tasks, including question answering and textual inference. The document outlines the model architecture of BERT and the pre-training and fine-tuning methods used.
Enhancing Entity Linking by Combining NER ModelsJulien PLU
The document describes enhancements made to the ADEL entity linking framework. ADEL combines multiple named entity recognition models and uses a combination of linguistic and dictionary-based approaches. New features in ADEL include using a generic API to interface with NLP tools, combining multiple CRF models for entity extraction, clustering nil entities, and developing a new backend using Elasticsearch and Couchbase. The document compares the performance of the original 2015 version of ADEL to the new 2016 version on standard entity linking tasks and datasets.
In this talk, you will discover how the 15k LOC codebase was implemented with spec so you don't have to (but probably should). Validation; testing; destructuring; composable “data macros” via conformers; we’ve tried spec in all its multifaceted glory. You will discover a distillation of lessons learned interspersed with musing on how spec alters development flow and one’s thinking.
Machine learning on graphs is an important and ubiquitous task with applications ranging from drug design to friendship recommendation in social networks. The primary challenge in this domain is finding a way to represent, or encode, graph structure so that it can be easily exploited by machine learning models. However, traditionally machine learning approaches relied on user-defined heuristics to extract features encoding structural information about a graph. In this talk I will discuss methods that automatically learn to encode graph structure into low-dimensional embeddings, using techniques based on deep learning and nonlinear dimensionality reduction. I will provide a conceptual review of key advancements in this area of representation learning on graphs, including random-walk based algorithms, and graph convolutional networks.
KOSMOS-1 is a multimodal large language model that can perceive and process language as well as visual inputs like images. It was trained on large web-scale datasets containing text, images, and image-caption pairs to align its vision capabilities with its natural language understanding. Experimental results showed that KOSMOS-1 can perform well on tasks involving language, vision, and their combination, including image captioning, visual question answering, and describing images based on text instructions, all without any fine-tuning. The ability to perceive and understand different modalities allows language models to acquire knowledge in new ways and expands their application to areas like robotics and document intelligence.
[Mmlab seminar 2016] deep learning for human pose estimationWei Yang
This document summarizes recent advances in deep learning approaches for human pose estimation. It describes early methods like DeepPose that used cascades of regressors. Later works introduced heatmap regression to capture spatial information. Convolutional Pose Machine and Stacked Hourglass networks further improved accuracy by incorporating stronger context modeling through deeper networks with larger receptive fields and intermediate supervision. These approaches demonstrate that both local appearance cues and modeling of global context and structure are important for accurate human pose estimation.
What Does the Webinar Cover?
You'll learn how to optimize varying parameters and disciplines throughout the lifecycle of the system within cost and schedule constraints without compromising performance. Real MBSE enables the execution of many activities in parallel, thus enabling the “faster and cheaper” part.
Many people can contribute to the design and development at the same time, because the information they create can be easily linked together to form abstractions that enable you to communicate the results at all levels. Dr. Dam uses a methodology that includes the technique, processes, and tools.
This methodology isn’t the only way to have a successful MBSE capability, but all three elements must be incorporated in any methodology you use. We offer this methodology as one that has proven successful over the past decade. It is based on methodologies used since the 1960s, but updated to the modern cloud computing, artificial intelligence age; that's now emerging toward the end of the second decade of the 21st Century.
Often people today work in a similar manner to how their grandparents worked in the 1960s, just with electronic tools instead of paper and pencil. Just creating a “model” doesn’t mean you are doing effective MBSE. This webinar will show you how to take MBSE into the 21st century.
Natural Language Understanding of Systems Engineering ArtifactsÁkos Horváth
This paper examines in close relation two fields of growing importance: model-based systems engineering (MBSE) and natural language processing (NLP). System models provide a structured description of engineering data, whose inherent semantics often remains hard to explore. Natural language understanding, (i.e., the machine analysis of texts produced by humans) an important field of NLP, focuses on semantic text comprehension but cannot directly account for structured information sources.
Similar to Text Processing Like Humans Do: Visually Attacking and Shielding NLP Systems (20)
For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2024/06/temporal-event-neural-networks-a-more-efficient-alternative-to-the-transformer-a-presentation-from-brainchip/
Chris Jones, Director of Product Management at BrainChip , presents the “Temporal Event Neural Networks: A More Efficient Alternative to the Transformer” tutorial at the May 2024 Embedded Vision Summit.
The expansion of AI services necessitates enhanced computational capabilities on edge devices. Temporal Event Neural Networks (TENNs), developed by BrainChip, represent a novel and highly efficient state-space network. TENNs demonstrate exceptional proficiency in handling multi-dimensional streaming data, facilitating advancements in object detection, action recognition, speech enhancement and language model/sequence generation. Through the utilization of polynomial-based continuous convolutions, TENNs streamline models, expedite training processes and significantly diminish memory requirements, achieving notable reductions of up to 50x in parameters and 5,000x in energy consumption compared to prevailing methodologies like transformers.
Integration with BrainChip’s Akida neuromorphic hardware IP further enhances TENNs’ capabilities, enabling the realization of highly capable, portable and passively cooled edge devices. This presentation delves into the technical innovations underlying TENNs, presents real-world benchmarks, and elucidates how this cutting-edge approach is positioned to revolutionize edge AI across diverse applications.
Have you ever been confused by the myriad of choices offered by AWS for hosting a website or an API?
Lambda, Elastic Beanstalk, Lightsail, Amplify, S3 (and more!) can each host websites + APIs. But which one should we choose?
Which one is cheapest? Which one is fastest? Which one will scale to meet our needs?
Join me in this session as we dive into each AWS hosting service to determine which one is best for your scenario and explain why!
"Choosing proper type of scaling", Olena SyrotaFwdays
Imagine an IoT processing system that is already quite mature and production-ready and for which client coverage is growing and scaling and performance aspects are life and death questions. The system has Redis, MongoDB, and stream processing based on ksqldb. In this talk, firstly, we will analyze scaling approaches and then select the proper ones for our system.
The Microsoft 365 Migration Tutorial For Beginner.pptxoperationspcvita
This presentation will help you understand the power of Microsoft 365. However, we have mentioned every productivity app included in Office 365. Additionally, we have suggested the migration situation related to Office 365 and how we can help you.
You can also read: https://www.systoolsgroup.com/updates/office-365-tenant-to-tenant-migration-step-by-step-complete-guide/
Discover top-tier mobile app development services, offering innovative solutions for iOS and Android. Enhance your business with custom, user-friendly mobile applications.
5th LF Energy Power Grid Model Meet-up SlidesDanBrown980551
5th Power Grid Model Meet-up
It is with great pleasure that we extend to you an invitation to the 5th Power Grid Model Meet-up, scheduled for 6th June 2024. This event will adopt a hybrid format, allowing participants to join us either through an online Mircosoft Teams session or in person at TU/e located at Den Dolech 2, Eindhoven, Netherlands. The meet-up will be hosted by Eindhoven University of Technology (TU/e), a research university specializing in engineering science & technology.
Power Grid Model
The global energy transition is placing new and unprecedented demands on Distribution System Operators (DSOs). Alongside upgrades to grid capacity, processes such as digitization, capacity optimization, and congestion management are becoming vital for delivering reliable services.
Power Grid Model is an open source project from Linux Foundation Energy and provides a calculation engine that is increasingly essential for DSOs. It offers a standards-based foundation enabling real-time power systems analysis, simulations of electrical power grids, and sophisticated what-if analysis. In addition, it enables in-depth studies and analysis of the electrical power grid’s behavior and performance. This comprehensive model incorporates essential factors such as power generation capacity, electrical losses, voltage levels, power flows, and system stability.
Power Grid Model is currently being applied in a wide variety of use cases, including grid planning, expansion, reliability, and congestion studies. It can also help in analyzing the impact of renewable energy integration, assessing the effects of disturbances or faults, and developing strategies for grid control and optimization.
What to expect
For the upcoming meetup we are organizing, we have an exciting lineup of activities planned:
-Insightful presentations covering two practical applications of the Power Grid Model.
-An update on the latest advancements in Power Grid -Model technology during the first and second quarters of 2024.
-An interactive brainstorming session to discuss and propose new feature requests.
-An opportunity to connect with fellow Power Grid Model enthusiasts and users.
"Frontline Battles with DDoS: Best practices and Lessons Learned", Igor IvaniukFwdays
At this talk we will discuss DDoS protection tools and best practices, discuss network architectures and what AWS has to offer. Also, we will look into one of the largest DDoS attacks on Ukrainian infrastructure that happened in February 2022. We'll see, what techniques helped to keep the web resources available for Ukrainians and how AWS improved DDoS protection for all customers based on Ukraine experience
HCL Notes und Domino Lizenzkostenreduzierung in der Welt von DLAUpanagenda
Webinar Recording: https://www.panagenda.com/webinars/hcl-notes-und-domino-lizenzkostenreduzierung-in-der-welt-von-dlau/
DLAU und die Lizenzen nach dem CCB- und CCX-Modell sind für viele in der HCL-Community seit letztem Jahr ein heißes Thema. Als Notes- oder Domino-Kunde haben Sie vielleicht mit unerwartet hohen Benutzerzahlen und Lizenzgebühren zu kämpfen. Sie fragen sich vielleicht, wie diese neue Art der Lizenzierung funktioniert und welchen Nutzen sie Ihnen bringt. Vor allem wollen Sie sicherlich Ihr Budget einhalten und Kosten sparen, wo immer möglich. Das verstehen wir und wir möchten Ihnen dabei helfen!
Wir erklären Ihnen, wie Sie häufige Konfigurationsprobleme lösen können, die dazu führen können, dass mehr Benutzer gezählt werden als nötig, und wie Sie überflüssige oder ungenutzte Konten identifizieren und entfernen können, um Geld zu sparen. Es gibt auch einige Ansätze, die zu unnötigen Ausgaben führen können, z. B. wenn ein Personendokument anstelle eines Mail-Ins für geteilte Mailboxen verwendet wird. Wir zeigen Ihnen solche Fälle und deren Lösungen. Und natürlich erklären wir Ihnen das neue Lizenzmodell.
Nehmen Sie an diesem Webinar teil, bei dem HCL-Ambassador Marc Thomas und Gastredner Franz Walder Ihnen diese neue Welt näherbringen. Es vermittelt Ihnen die Tools und das Know-how, um den Überblick zu bewahren. Sie werden in der Lage sein, Ihre Kosten durch eine optimierte Domino-Konfiguration zu reduzieren und auch in Zukunft gering zu halten.
Diese Themen werden behandelt
- Reduzierung der Lizenzkosten durch Auffinden und Beheben von Fehlkonfigurationen und überflüssigen Konten
- Wie funktionieren CCB- und CCX-Lizenzen wirklich?
- Verstehen des DLAU-Tools und wie man es am besten nutzt
- Tipps für häufige Problembereiche, wie z. B. Team-Postfächer, Funktions-/Testbenutzer usw.
- Praxisbeispiele und Best Practices zum sofortigen Umsetzen
Northern Engraving | Nameplate Manufacturing Process - 2024Northern Engraving
Manufacturing custom quality metal nameplates and badges involves several standard operations. Processes include sheet prep, lithography, screening, coating, punch press and inspection. All decoration is completed in the flat sheet with adhesive and tooling operations following. The possibilities for creating unique durable nameplates are endless. How will you create your brand identity? We can help!
Taking AI to the Next Level in Manufacturing.pdfssuserfac0301
Read Taking AI to the Next Level in Manufacturing to gain insights on AI adoption in the manufacturing industry, such as:
1. How quickly AI is being implemented in manufacturing.
2. Which barriers stand in the way of AI adoption.
3. How data quality and governance form the backbone of AI.
4. Organizational processes and structures that may inhibit effective AI adoption.
6. Ideas and approaches to help build your organization's AI strategy.
Introduction of Cybersecurity with OSS at Code Europe 2024Hiroshi SHIBATA
I develop the Ruby programming language, RubyGems, and Bundler, which are package managers for Ruby. Today, I will introduce how to enhance the security of your application using open-source software (OSS) examples from Ruby and RubyGems.
The first topic is CVE (Common Vulnerabilities and Exposures). I have published CVEs many times. But what exactly is a CVE? I'll provide a basic understanding of CVEs and explain how to detect and handle vulnerabilities in OSS.
Next, let's discuss package managers. Package managers play a critical role in the OSS ecosystem. I'll explain how to manage library dependencies in your application.
I'll share insights into how the Ruby and RubyGems core team works to keep our ecosystem safe. By the end of this talk, you'll have a better understanding of how to safeguard your code.
Freshworks Rethinks NoSQL for Rapid Scaling & Cost-EfficiencyScyllaDB
Freshworks creates AI-boosted business software that helps employees work more efficiently and effectively. Managing data across multiple RDBMS and NoSQL databases was already a challenge at their current scale. To prepare for 10X growth, they knew it was time to rethink their database strategy. Learn how they architected a solution that would simplify scaling while keeping costs under control.
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectorsDianaGray10
Join us to learn how UiPath Apps can directly and easily interact with prebuilt connectors via Integration Service--including Salesforce, ServiceNow, Open GenAI, and more.
The best part is you can achieve this without building a custom workflow! Say goodbye to the hassle of using separate automations to call APIs. By seamlessly integrating within App Studio, you can now easily streamline your workflow, while gaining direct access to our Connector Catalog of popular applications.
We’ll discuss and demo the benefits of UiPath Apps and connectors including:
Creating a compelling user experience for any software, without the limitations of APIs.
Accelerating the app creation process, saving time and effort
Enjoying high-performance CRUD (create, read, update, delete) operations, for
seamless data management.
Speakers:
Russell Alfeche, Technology Leader, RPA at qBotic and UiPath MVP
Charlie Greenberg, host
Programming Foundation Models with DSPy - Meetup SlidesZilliz
Prompting language models is hard, while programming language models is easy. In this talk, I will discuss the state-of-the-art framework DSPy for programming foundation models with its powerful optimizers and runtime constraint system.
Your One-Stop Shop for Python Success: Top 10 US Python Development Providersakankshawande
Simplify your search for a reliable Python development partner! This list presents the top 10 trusted US providers offering comprehensive Python development services, ensuring your project's success from conception to completion.
Your One-Stop Shop for Python Success: Top 10 US Python Development Providers
Text Processing Like Humans Do: Visually Attacking and Shielding NLP Systems
1. Paper information
1
• Title
ü Text Processing Like Humans Do: Visually Attacking and
Shielding NLP Systems
• URL
ü https://aclweb.org/anthology/papers/N/N19/N19-1165/
• Author
ü Steffen Eger, Gözde Gül Şahin, Andreas Rücklé, Ji-Ung Lee,
Claudia Schulz, Mohsen Mesgar, Krishnkant Swarnkar,
Edwin Simpson, Iryna Gurevych
• Conference
ü NAACL2019
2. Background: visual perturbations to text
2
• Visual perturbations to text are often used to
obfuscate offensive comments in social media
• Those perturbations are considered as a new type of
adversarial attack in NLP
1 4M JUST GO1NG TO K1LL YOU ƒv¢K !!
You are f**ck!ng !d!0t
Adversarial attack:
Make modifications to an input to fool the system, while
the original meaning is still understood by humans
3. Background:
Advantages of visual perturbations
3
1. They do not require any linguistic knowledge beyond the
character level
2. They are less damaging to human perception than syntax
errors or the insertion of nagations
3. They do not require knowledge of the attacked model
In summary, visual perturbations are easily
applicable to any languages, domains and tasks
Perturbed: 1 4M JUST GO1NG TO K1LL YOU ƒv¢K !!
⇅
Raw: I AM JUST GOING TO KILL YOU FUCK !!
4. Summary of this paper:
4
• Develop three methods for visual perturbations
• Confirm that humans are robust to visual perturbations
• Confirm that the performance of SOTA NLP models
drops when attacked by visual perturbations
• Develop three methods to shield from visual attacks
5. Summary of this paper:
5
• Develop three methods for visual perturbations
• Confirm that humans are robust to visual perturbations
• Confirm that the performance of SOTA NLP models
drops when attacked by visual perturbations
• Develop three methods to shield from visual attacks
6. Proposed visual perturbations
6
Proposed methods perturb input sentences by
replacing each character randomly based on:
• Image-based character embedding (ICES)
• Description-based character embedding (DCES)
• Easy-character embedding (ECES)
7. 7
Image-based character embedding (ICES)
ü retrieve a 24*24 image of the character and convert it into
576 dimensional embedding vector
ü replace characters of the input sentences by their nearest
neighbors in the embedding space
Proposed visual perturbations:
Image-based
c
ć
Ҫ
ą
ă
a
embedding
space
8. 8
Description-based character embedding (DCES)
ü retrieve the description of each Unicode character
ü replace characters by other ones whose description shares
many of the words of the target description
a - latin small letter “a”
à - latin small letter “a” with grave
description:
replace
Proposed visual perturbations:
Descriptions-based
9. 9
Easy-character-based character embedding (ECES)
ü replace characters of the input sentences by manually
defined characters (targets are 52 characters: a-zA-Z)
a → â
b → ḃ
c → ĉ
:
rule: replace
Proposed visual perturbations:
Easy-character-based
manually
defined
10. • Ten nearest neighbors in different character spaces
• Examples of perturbed and original sentences
Proposed visual perturbations:
Easy-character-based
10
ECES-0.8
flipping probability of perturbations
11. Summary of this paper:
11
• Develop three methods for visual perturbations
• Confirm that humans are robust to visual
perturbations
• Confirm that the performance of SOTA NLP models
drops when attacked by visual perturbations
• Develop three methods to shield from visual attacks
12. 12
To evaluate human performances, asked annotators to
recover the original sentences given perturbed text
ü calculate error rate by measuring the normalized edit distance
between the recovered sentence and the original one
Human annotation experiment against
visual perturbation
Flipping probability p
Errorratein%
Humans are very good at understanding visual perturbationsbetter
ECES
13. Summary of this paper:
13
• Develop three methods for visual perturbations
• Confirm that humans are robust to visual perturbations
• Confirm that the performance of SOTA NLP models
drops when attacked by visual perturbations
• Develop three methods to shield from visual attacks
14. 14
Evaluate the capabilities of SOTA NLP models for
below tasks to deal with visual attacks (by DCES)
• POS tagging (POS)
• Chunking (Chunk)
ü Dataset: CoNLL 2000
ü Model: Bi-LSTM with ELMo
• Grapheme-to-phoneme (G2P)
ü Dataset: Combilex pronunciation of American English
ü Model: Bi-LSTM
• Toxic comment classification (TC)
ü Dataset: Kaggle dataset
ü Model: Feed-forward network with ELMo
Computational experiment against
visual perturbation: settings
15. 15
Show the relative performance s*(p) compared to
the performance of no perturbations s(0)
Computational experiment against
visual perturbation (no shielding)
better
= s*(p)
All systems degrade considerably compared
to the systems with no perturbations
p
16. Summary of this paper:
16
• Develop three methods for visual perturbations
• Confirm that humans are robust to visual perturbations
• Confirm that the performance of SOTA NLP models
drops when attacked by visual perturbations
• Develop three methods to shield from visual attacks
17. 17
Develop three shielding methods against visual attacks
• Adversarial training (AT)
ü Replace original training examples by perturbed data
• Visual character embedding (CE)
ü Use fixed ICEs to initialize the embeddings of the models
• Rule-based recovery (RBR)
ü Replace each non-standard character in the input with its
nearest standard neighbor in ICES (a-zA-Z + punctuation)
Proposed shielding methods against
visual perturbations
18. 18
Show the performance improvements Δ between
shielding treatments σ(p)/s(0) and original scores s*(p)
Proposed shielding methods against
visual perturbations: results
better
ΔAT
ΔCE
ΔAT+CE
ΔRBR
p p
= (p)/s(0) s(p)/s(0)<latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit>
19. 19
Proposed shielding methods against
visual perturbations: results
better
ΔAT
ΔCE
ΔAT+CE
ΔRBR
p p
All tasks other than G2P profit from AT
• AT did not perform well on G2P because missing
tokens are more problematic than other tasks
= (p)/s(0) s(p)/s(0)<latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">AAAChnichVFNS1tBFD0+q8b4FXUjuEkNSlyY3ieKRSiE6qJLv6KCkfDe6xgH3xdvJgEbui70D7hwVUFEpNv6A9z0D3ThTyguFdy48ObliVip3mFmzpy5586ZGTt0pdJEl21G+5uOzq5Ud7qnt69/IDM4tK6CWuSIkhO4QbRpW0q40hclLbUrNsNIWJ7tig17b6G5v1EXkZKBv6b3Q7HtWVVf7kjH0kxVMm/Li8LV1oeyklXPyoeT71SeJrNTWfWAK5kcFSiO7HNgJiCHJJaCzAnK+IwADmrwIOBDM3ZhQXHbgglCyNw2GsxFjGS8L/AVadbWOEtwhsXsHo9VXm0lrM/rZk0Vqx0+xeUesTKLcfpDp3RNv+mM/tLdf2s14hpNL/s82y2tCCsD30dWb19VeTxr7D6qXvSssYP3sVfJ3sOYad7CaenrXw6uV+dXxhsTdERX7P8HXdIF38Cv3zjHy2LlEGn+APPf534O1qcLJhXM5Zlc8WPyFSmMYgx5fu85FPEJSyjxud/wE79wbqSMgjFrzLVSjbZEM4wnYRTvAfXek4o=</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit>
20. 20
Proposed shielding methods against
visual perturbations: results
better
ΔAT
ΔCE
ΔAT+CE
ΔRBR
p p
TC and G2P profit from CE
• CE can restore tokens from those neighborhoods in the
embedding space
• CE did not perform well on POS and Chunk because ELMo
might weaken the effect of CE
= (p)/s(0) s(p)/s(0)<latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">AAAChnichVFNS1tBFD0+q8b4FXUjuEkNSlyY3ieKRSiE6qJLv6KCkfDe6xgH3xdvJgEbui70D7hwVUFEpNv6A9z0D3ThTyguFdy48ObliVip3mFmzpy5586ZGTt0pdJEl21G+5uOzq5Ud7qnt69/IDM4tK6CWuSIkhO4QbRpW0q40hclLbUrNsNIWJ7tig17b6G5v1EXkZKBv6b3Q7HtWVVf7kjH0kxVMm/Li8LV1oeyklXPyoeT71SeJrNTWfWAK5kcFSiO7HNgJiCHJJaCzAnK+IwADmrwIOBDM3ZhQXHbgglCyNw2GsxFjGS8L/AVadbWOEtwhsXsHo9VXm0lrM/rZk0Vqx0+xeUesTKLcfpDp3RNv+mM/tLdf2s14hpNL/s82y2tCCsD30dWb19VeTxr7D6qXvSssYP3sVfJ3sOYad7CaenrXw6uV+dXxhsTdERX7P8HXdIF38Cv3zjHy2LlEGn+APPf534O1qcLJhXM5Zlc8WPyFSmMYgx5fu85FPEJSyjxud/wE79wbqSMgjFrzLVSjbZEM4wnYRTvAfXek4o=</latexit>
21. 21
Proposed shielding methods against
visual perturbations: results
better
ΔAT
ΔCE
ΔAT+CE
ΔRBR
p p
All tasks profit from AT with CE
• The combination of them can boost the effect of each other
= (p)/s(0) s(p)/s(0)<latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">AAAChnichVFNS1tBFD0+q8b4FXUjuEkNSlyY3ieKRSiE6qJLv6KCkfDe6xgH3xdvJgEbui70D7hwVUFEpNv6A9z0D3ThTyguFdy48ObliVip3mFmzpy5586ZGTt0pdJEl21G+5uOzq5Ud7qnt69/IDM4tK6CWuSIkhO4QbRpW0q40hclLbUrNsNIWJ7tig17b6G5v1EXkZKBv6b3Q7HtWVVf7kjH0kxVMm/Li8LV1oeyklXPyoeT71SeJrNTWfWAK5kcFSiO7HNgJiCHJJaCzAnK+IwADmrwIOBDM3ZhQXHbgglCyNw2GsxFjGS8L/AVadbWOEtwhsXsHo9VXm0lrM/rZk0Vqx0+xeUesTKLcfpDp3RNv+mM/tLdf2s14hpNL/s82y2tCCsD30dWb19VeTxr7D6qXvSssYP3sVfJ3sOYad7CaenrXw6uV+dXxhsTdERX7P8HXdIF38Cv3zjHy2LlEGn+APPf534O1qcLJhXM5Zlc8WPyFSmMYgx5fu85FPEJSyjxud/wE79wbqSMgjFrzLVSjbZEM4wnYRTvAfXek4o=</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit>
22. 22
Proposed shielding methods against
visual perturbations: results
better
ΔAT
ΔCE
ΔAT+CE
ΔRBR
p p
All tasks profit from RBR lower than AT + CE
• RBR may incorrectly replace input tokens that affect the
performances
= (p)/s(0) s(p)/s(0)<latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">AAAChnichVFNS1tBFD0+q8b4FXUjuEkNSlyY3ieKRSiE6qJLv6KCkfDe6xgH3xdvJgEbui70D7hwVUFEpNv6A9z0D3ThTyguFdy48ObliVip3mFmzpy5586ZGTt0pdJEl21G+5uOzq5Ud7qnt69/IDM4tK6CWuSIkhO4QbRpW0q40hclLbUrNsNIWJ7tig17b6G5v1EXkZKBv6b3Q7HtWVVf7kjH0kxVMm/Li8LV1oeyklXPyoeT71SeJrNTWfWAK5kcFSiO7HNgJiCHJJaCzAnK+IwADmrwIOBDM3ZhQXHbgglCyNw2GsxFjGS8L/AVadbWOEtwhsXsHo9VXm0lrM/rZk0Vqx0+xeUesTKLcfpDp3RNv+mM/tLdf2s14hpNL/s82y2tCCsD30dWb19VeTxr7D6qXvSssYP3sVfJ3sOYad7CaenrXw6uV+dXxhsTdERX7P8HXdIF38Cv3zjHy2LlEGn+APPf534O1qcLJhXM5Zlc8WPyFSmMYgx5fu85FPEJSyjxud/wE79wbqSMgjFrzLVSjbZEM4wnYRTvAfXek4o=</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit><latexit sha1_base64="y0dGgyoE1b/V7cP6jkz5sSEf7iE=">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</latexit>
23. 23
Show examples of the prediction in TC (flipping prob. = 0.1)
Proposed shielding methods against
visual perturbations: example
ECES
DCES
ECES
DCES
• Perturbing specific words reduces the score of a non-shielded
approach, while perturbing useless words like ‘he’ has little effect
Answer Prediction
24. 24
Show examples of the prediction in TC (flipping prob. = 0.1)
Proposed shielding methods against
visual perturbations: example
ECES
DCES
ECES
DCES
• Perturbing specific words reduces the score of a non-shielded
approach, while perturbing useless words like ‘he’ has little effect
• Overall, all the shielding approaches help in various degrees
Answer Prediction
25. Summary of this paper:
25
• Develop three methods for visual perturbations
• Confirm that humans are robust to visual perturbations
• Confirm that the performance of SOTA NLP models
drops when attacked by visual perturbations
• Develop three methods to shield from visual attacks