Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userHow Do Neural Sequence Models Generalize? Local and Global Cues for Out-of-Distribution Prediction
كيف تعميم نماذج التسلسل العصبي؟العظة المحلية والعالمية للتنبؤ خارج التوزيع
372
0 0
0.0
(
0
)
Created by
Shamra Editor
Ask ChatGPT about the research
بعد أن يواجه نموذج التسلسل العصبي رمزية غير متوقعة، هل يمكن التنبؤ بسلوكه؟ نظهر أن نماذج Language RNN وحول المحولات تعرض تعميم مهيكلا متسقا في سياقات خارج التوزيع. نبدأ بإدخال نماذجين مثالية من التعميم في التنبؤ التالي بالكلمة التالية: نموذج سياق معجمي يعمل فيه التعميم يتفق مع الكلمة الأخيرة الملاحظة، ونموذج السياق النحامي الذي يتوافق فيه التعميم مع الهيكل العالمي للمدخلات. في تجارب باللغة الإنجليزية والفنلندية والماندرين ولغات منتظمة عشوائية، نوضح أن نماذج اللغة العصبية محبط بين هذين الشكلين من التعميم: تنبؤاتها تقارب جيدا من خلال مزيج خطي من التوزيعات التنبؤية المعجمية والنوعية. ثم نوضح ذلك، في بعض اللغات، يتوسط الضوضاء شكلين التعميم: الضوضاء المطبقة على رموز المدخلات تشجع التعميم النحوي، في حين أن الضوضاء في تمثيلات التاريخ تشجع التعميم المعجمي. أخيرا، نقدم شرحا نظريا أوليا لهذه النتائج من خلال إثبات أن سلوك الاستيفاء الملحوظ متوقع في النماذج الخطية من السجل مع هيكل ارتباط ميزة معينة. تساعد هذه النتائج في تفسير فعالية خططيتين تنظيمي شعبيتين وإظهار أن جوانب تعميم نموذج التسلسل يمكن فهمها والسيطر عليها.
After a neural sequence model encounters an unexpected token, can its behavior be predicted? We show that RNN and transformer language models exhibit structured, consistent generalization in out-of-distribution contexts. We begin by introducing two idealized models of generalization in next-word prediction: a lexical context model in which generalization is consistent with the last word observed, and a syntactic context model in which generalization is consistent with the global structure of the input. In experiments in English, Finnish, Mandarin, and random regular languages, we demonstrate that neural language models interpolate between these two forms of generalization: their predictions are well-approximated by a log-linear combination of lexical and syntactic predictive distributions. We then show that, in some languages, noise mediates the two forms of generalization: noise applied to input tokens encourages syntactic generalization, while noise in history representations encourages lexical generalization. Finally, we offer a preliminary theoretical explanation of these results by proving that the observed interpolation behavior is expected in log-linear models with a particular feature correlation structure. These results help explain the effectiveness of two popular regularization schemes and show that aspects of sequence model generalization can be understood and controlled.
References used
https://aclanthology.org/
rate research
Read More
Large pretrained language models using the transformer neural network architecture are becoming a dominant methodology for many natural language processing tasks, such as question answering, text classification, word sense disambiguation, text comple
tion and machine translation. Commonly comprising hundreds of millions of parameters, these models offer state-of-the-art performance, but at the expense of interpretability. The attention mechanism is the main component of transformer networks. We present AttViz, a method for exploration of self-attention in transformer networks, which can help in explanation and debugging of the trained models by showing associations between text tokens in an input sequence. We show that existing deep learning pipelines can be explored with AttViz, which offers novel visualizations of the attention heads and their aggregations. We implemented the proposed methods in an online toolkit and an offline library. Using examples from news analysis, we demonstrate how AttViz can be used to inspect and potentially better understand what a model has learned.
Recent studies have shown that a bias in thetext suggestions system can percolate in theuser's writing. In this pilot study, we ask thequestion: How do people interact with text pre-diction models, in an inline next phrase sugges-tion interface and h
ow does introducing senti-ment bias in the text prediction model affecttheir writing? We present a pilot study as afirst step to answer this question.
Pretrained Transformers achieve remarkable performance when training and test data are from the same distribution. However, in real-world scenarios, the model often faces out-of-distribution (OOD) instances that can cause severe semantic shift proble
ms at inference time. Therefore, in practice, a reliable model should identify such instances, and then either reject them during inference or pass them over to models that handle another distribution. In this paper, we develop an unsupervised OOD detection method, in which only the in-distribution (ID) data are used in training. We propose to fine-tune the Transformers with a contrastive loss, which improves the compactness of representations, such that OOD instances can be better differentiated from ID ones. These OOD instances can then be accurately detected using the Mahalanobis distance in the model's penultimate layer. We experiment with comprehensive settings and achieve near-perfect OOD detection performance, outperforming baselines drastically. We further investigate the rationales behind the improvement, finding that more compact representations through margin-based contrastive learning bring the improvement. We release our code to the community for future research.
While neural networks are ubiquitous in state-of-the-art semantic parsers, it has been shown that most standard models suffer from dramatic performance losses when faced with compositionally out-of-distribution (OOD) data. Recently several methods ha
ve been proposed to improve compositional generalization in semantic parsing. In this work we instead focus on the problem of detecting compositionally OOD examples with neural semantic parsers, which, to the best of our knowledge, has not been investigated before. We investigate several strong yet simple methods for OOD detection based on predictive uncertainty. The experimental results demonstrate that these techniques perform well on the standard SCAN and CFQ datasets. Moreover, we show that OOD detection can be further improved by using a heterogeneous ensemble.
Natural language (NL) explanations of model predictions are gaining popularity as a means to understand and verify decisions made by large black-box pre-trained models, for tasks such as Question Answering (QA) and Fact Verification. Recently, pre-tr
ained sequence to sequence (seq2seq) models have proven to be very effective in jointly making predictions, as well as generating NL explanations. However, these models have many shortcomings; they can fabricate explanations even for incorrect predictions, they are difficult to adapt to long input documents, and their training requires a large amount of labeled data. In this paper, we develop FiD-Ex, which addresses these shortcomings for seq2seq models by: 1) introducing sentence markers to eliminate explanation fabrication by encouraging extractive generation, 2) using the fusion-in-decoder architecture to handle long input contexts, and 3) intermediate fine-tuning on re-structured open domain QA datasets to improve few-shot performance. FiD-Ex significantly improves over prior work in terms of explanation metrics and task accuracy on five tasks from the ERASER explainability benchmark in both fully supervised and few-shot settings.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
