机器学习中的终身学习范式是一个有吸引力的替代方案，不仅是由于其与生物学学习的相似之处，而且它通过避免过度模型重新训练来减少能量浪费的可能性。对此范式的关键挑战是灾难性遗忘的现象。随着在机器学习中训练有素的模型的越来越受欢迎和成功，我们提出了问题：终身学习中的训练前比赛，特别是关于灾难性的遗忘？我们在大型预先训练模型的上下文中调查现有方法，并在各种文本和图像分类任务中评估其性能，包括使用15个不同的NLP任务的新型数据集进行大规模研究。在所有设置中，我们观察到，通用预训练隐含地减轻了在与随机初始化模型相比依次学习多个任务时灾难性忘记的影响。然后，我们进一步调查为什么预先训练缓解在这个环境中忘记。我们通过分析损失景观来研究这种现象，发现预先训练的重量似乎可以通过导致更宽的最小值来缓解遗忘。基于这一洞察力，我们提出了对当前任务损失和损失盆地锐利的共同优化，以便在连续微调期间明确鼓励更广泛的盆地。我们表明，这种优化方法导致与跨多个设置的任务顺序持续学习的性能相当，而无需保留具有任务数量的大小的内存。

Differentiable Search Indices (DSIs) encode a corpus of documents in the parameters of a model and use the same model to map queries directly to relevant document identifiers. Despite the strong performance of DSI models, deploying them in situations where the corpus changes over time is computationally expensive because reindexing the corpus requires re-training the model. In this work, we introduce DSI++, a continual learning challenge for DSI to incrementally index new documents while being able to answer queries related to both previously and newly indexed documents. Across different model scales and document identifier representations, we show that continual indexing of new documents leads to considerable forgetting of previously indexed documents. We also hypothesize and verify that the model experiences forgetting events during training, leading to unstable learning. To mitigate these issues, we investigate two approaches. The first focuses on modifying the training dynamics. Flatter minima implicitly alleviate forgetting, so we optimize for flatter loss basins and show that the model stably memorizes more documents (+12\%). Next, we introduce a generative memory to sample pseudo-queries for documents and supplement them during continual indexing to prevent forgetting for the retrieval task. Extensive experiments on novel continual indexing benchmarks based on Natural Questions (NQ) and MS MARCO demonstrate that our proposed solution mitigates forgetting by a significant margin. Concretely, it improves the average Hits@10 by $+21.1\%$ over competitive baselines for NQ and requires $6$ times fewer model updates compared to re-training the DSI model for incrementally indexing five corpora in a sequence.

持续学习研究的主要重点领域是通过设计新算法对分布变化更强大的新算法来减轻神经网络中的“灾难性遗忘”问题。尽管持续学习文献的最新进展令人鼓舞，但我们对神经网络的特性有助于灾难性遗忘的理解仍然有限。为了解决这个问题，我们不关注持续的学习算法，而是在这项工作中专注于模型本身，并研究神经网络体系结构对灾难性遗忘的“宽度”的影响，并表明宽度在遗忘遗产方面具有出人意料的显着影响。为了解释这种效果，我们从各个角度研究网络的学习动力学，例如梯度正交性，稀疏性和懒惰的培训制度。我们提供了与不同架构和持续学习基准之间的经验结果一致的潜在解释。

持续学习背后的主流范例一直在使模型参数调整到非静止数据分布，灾难性遗忘是中央挑战。典型方法在测试时间依赖排练缓冲区或已知的任务标识，以检索学到的知识和地址遗忘，而这项工作呈现了一个新的范例，用于持续学习，旨在训练更加简洁的内存系统而不在测试时间访问任务标识。我们的方法学会动态提示（L2P）预先训练的模型，以在不同的任务转换下顺序地学习任务。在我们提出的框架中，提示是小型可学习参数，这些参数在内存空间中保持。目标是优化提示，以指示模型预测并明确地管理任务不变和任务特定知识，同时保持模型可塑性。我们在流行的图像分类基准下进行全面的实验，具有不同挑战的持续学习环境，其中L2P始终如一地优于现有最先进的方法。令人惊讶的是，即使没有排练缓冲区，L2P即使没有排练缓冲，L2P也能实现竞争力的结果，并直接适用于具有挑战性的任务不可行的持续学习。源代码在https://github.com/google-Research/l2p中获得。

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a;Radford et al., 2018), BERT is designed to pretrain deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be finetuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial taskspecific architecture modifications.BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

在许多图像分类任务中，诸如夹子之类的开放式摄影模型具有高精度。但是，在某些设置中，他们的零拍摄性能远非最佳。我们研究模型修补程序，目的是提高对特定任务的准确性，而不会在表现已经足够的任务上降低准确性。为了实现这一目标，我们引入了油漆，这是一种修补方法，该方法在微调之前使用模型的权重与要修补的任务进行微调后的权重。在零机夹的性能差的九个任务上，油漆可将精度提高15至60个百分点，同时将ImageNet上的精度保留在零拍模型的一个百分点之内。油漆还允许在多个任务上修补单个模型，并通过模型刻度进行改进。此外，我们确定了广泛转移的案例，即使任务不相交，对一个任务进行修补也会提高其他任务的准确性。最后，我们研究了超出常见基准的应用程序，例如计数或减少印刷攻击对剪辑的影响。我们的发现表明，可以扩展一组任务集，开放式摄影模型可实现高精度，而无需从头开始重新训练它们。

持续学习旨在使单个模型能够学习一系列任务，而不会造成灾难性的遗忘。表现最好的方法通常需要排练缓冲区来存储过去的原始示例以进行经验重播，但是，由于隐私和内存约束，这会限制其实际价值。在这项工作中，我们提出了一个简单而有效的框架，即DualPrompt，该框架学习了一组称为提示的参数，以正确指示预先训练的模型，以依次学习到达的任务，而不会缓冲过去的示例。 DualPrompt提出了一种新颖的方法，可以将互补提示附加到预训练的主链上，然后将目标提出为学习任务不变和特定于任务的“指令”。通过广泛的实验验证，双启示始终在具有挑战性的课堂开发环境下始终设置最先进的表现。尤其是，双启示的表现优于最近的高级持续学习方法，其缓冲尺寸相对较大。我们还引入了一个更具挑战性的基准Split Imagenet-R，以帮助概括无连续的持续学习研究。源代码可在https://github.com/google-research/l2p上找到。

Artificial neural networks thrive in solving the classification problem for a particular rigid task, acquiring knowledge through generalized learning behaviour from a distinct training phase. The resulting network resembles a static entity of knowledge, with endeavours to extend this knowledge without targeting the original task resulting in a catastrophic forgetting. Continual learning shifts this paradigm towards networks that can continually accumulate knowledge over different tasks without the need to retrain from scratch. We focus on task incremental classification, where tasks arrive sequentially and are delineated by clear boundaries. Our main contributions concern (1) a taxonomy and extensive overview of the state-of-the-art; (2) a novel framework to continually determine the stability-plasticity trade-off of the continual learner; (3) a comprehensive experimental comparison of 11 state-of-the-art continual learning methods and 4 baselines. We empirically scrutinize method strengths and weaknesses on three benchmarks, considering Tiny Imagenet and large-scale unbalanced iNaturalist and a sequence of recognition datasets. We study the influence of model capacity, weight decay and dropout regularization, and the order in which the tasks are presented, and qualitatively compare methods in terms of required memory, computation time and storage.

在单独或多任务设置中评估了当前最新的视觉和语言模型，从而忽略了持续学习（CL）任务到达时的挑战。现有的CLENG分类促进了有关调整任务和减轻“灾难性遗忘”的研究，但仅限于仅视觉和仅语言的任务。我们提出了攀登，这是研究CL设置中学习多模式任务的挑战的基准，并系统地评估上游持续学习如何迅速概括为新的多模式和单峰任务。攀登包括几种CL算法的实现以及可以在多模式和单峰任务上部署的修改视觉语言变压器（VILT）模型。我们发现，常见的CL方法可以帮助减轻多模式任务学习期间的遗忘，但不要实现交叉任务知识转移。我们设想，攀登将有助于针对这种具有挑战性的多模式环境的新的CL算法进行研究。

Changing how pre-trained models behave -- e.g., improving their performance on a downstream task or mitigating biases learned during pre-training -- is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around \textit{task vectors}. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form ``A is to B as C is to D", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models.

A growing body of research in continual learning focuses on the catastrophic forgetting problem. While many attempts have been made to alleviate this problem, the majority of the methods assume a single model in the continual learning setup. In this work, we question this assumption and show that employing ensemble models can be a simple yet effective method to improve continual performance. However, ensembles' training and inference costs can increase significantly as the number of models grows. Motivated by this limitation, we study different ensemble models to understand their benefits and drawbacks in continual learning scenarios. Finally, to overcome the high compute cost of ensembles, we leverage recent advances in neural network subspace to propose a computationally cheap algorithm with similar runtime to a single model yet enjoying the performance benefits of ensembles.

Continual Learning (CL) is a field dedicated to devise algorithms able to achieve lifelong learning. Overcoming the knowledge disruption of previously acquired concepts, a drawback affecting deep learning models and that goes by the name of catastrophic forgetting, is a hard challenge. Currently, deep learning methods can attain impressive results when the data modeled does not undergo a considerable distributional shift in subsequent learning sessions, but whenever we expose such systems to this incremental setting, performance drop very quickly. Overcoming this limitation is fundamental as it would allow us to build truly intelligent systems showing stability and plasticity. Secondly, it would allow us to overcome the onerous limitation of retraining these architectures from scratch with the new updated data. In this thesis, we tackle the problem from multiple directions. In a first study, we show that in rehearsal-based techniques (systems that use memory buffer), the quantity of data stored in the rehearsal buffer is a more important factor over the quality of the data. Secondly, we propose one of the early works of incremental learning on ViTs architectures, comparing functional, weight and attention regularization approaches and propose effective novel a novel asymmetric loss. At the end we conclude with a study on pretraining and how it affects the performance in Continual Learning, raising some questions about the effective progression of the field. We then conclude with some future directions and closing remarks.

State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP.

经过审计的语言模型（PTLMS）通常是通过大型静态语料库学习的，并针对各种下游任务进行了微调。但是，当部署在现实世界中时，基于PTLM的模型必须处理偏离PTLM最初培训的数据分布。在本文中，我们研究了一个终身语言模型预处理挑战，其中不断更新PTLM以适应新兴数据。在域内收入的研究纸流和按时间顺序排序的推文流上，我们从具有不同持续学习算法的PTLM逐渐预处理PTLM，并跟踪下游任务性能（经过微调之后）。我们评估了PTLM在保留早期语料库中学习知识的同时适应新语料库的能力。我们的实验表明，基于蒸馏的方法最有效地在早期域中保持下游性能。该算法还可以改善知识传递，从而使模型能够比最新数据实现更好的下游性能，并在由于时间而在培训和评估之间存在分配差距时改善时间概括。我们认为，我们的问题制定，方法和分析将激发未来的研究朝着语言模型的持续预处理。

先前的关于自我监督预训练的研究重点是联合培训方案，在该场景中，假定大量未标记的数据一次性地将其作为输入，只有那时才受过培训的学习者。不幸的是，这种问题设置通常是不切实际的，即使不是不可行的，因为许多现实世界的任务依赖于顺序学习，例如，数据是以流方式分散或收集的。在本文中，我们对通过流数据进行了对自我监督的预训练进行了首次彻底而专门的研究，旨在阐明这种被忽视的设置下的模型行为。具体而言，我们在来自ImageNet和域内的四类预训练流数据数据上预先培训超过500个模型，并在三种类型的下游任务和12个不同的下游数据集上对其进行评估。我们的研究表明，以某种方式超出了我们的期望，通过简单的数据重播或参数正则化，顺序的自我监督预训练的预训练证明是联合预训练的有效替代方法，因为前者的性能主要与这些培训相同后者。此外，灾难性的遗忘是顺序监督学习中的一个常见问题，在顺序的自学学习（SSL）中得到了极大的缓解，这是通过我们对损失景观中最小值的表示和敏锐度的全面经验分析来很好地证明的。因此，我们的发现表明，在实践中，对于SSL，可以主要通过顺序学习来代替繁琐的联合培训，这反过来又可以更广泛的潜在应用方案。

Transfer learning, where a model is first pre-trained on a data-rich task before being finetuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts all text-based language problems into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled data sets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new "Colossal Clean Crawled Corpus", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our data set, pre-trained models, and code.

随着时间的流逝，不断扩大知识并利用其快速推广到新任务的能力是人类语言智能的关键特征。然而，现有对新任务进行快速概括的模型（例如，很少的学习方法）主要是在固定数据集中的单个镜头中训练，无法动态扩展其知识；虽然不断学习算法并非专门设计用于快速概括。我们提出了一种新的学习设置，对几杆学习者（CLIF）的持续学习，以应对统一设置的两个学习设置的挑战。 CLIF假设模型从依次到达的一系列不同的NLP任务中学习，从而积累了知识，以改善对新任务的概括，同时还保留了较早所学的任务的性能。我们研究了在持续学习设置中如何影响概括能力，评估许多持续学习算法，并提出一种新型的正则适配器生成方法。我们发现，灾难性的遗忘影响着概括能力的程度远低于所见任务的表现。虽然持续学习算法仍然可以为概括能力带来可观的好处。

Language model pre-training has proven to be useful in learning universal language representations. As a state-of-the-art language model pre-training model, BERT (Bidirectional Encoder Representations from Transformers) has achieved amazing results in many language understanding tasks. In this paper, we conduct exhaustive experiments to investigate different fine-tuning methods of BERT on text classification task and provide a general solution for BERT fine-tuning. Finally, the proposed solution obtains new state-of-the-art results on eight widely-studied text classification datasets. 1

当代理在终身学习设置中遇到连续的新任务流时，它利用了从早期任务中获得的知识来帮助更好地学习新任务。在这种情况下，确定有效的知识表示成为一个具有挑战性的问题。大多数研究工作都建议将过去任务中的一部分示例存储在重播缓冲区中，将一组参数集成给每个任务，或通过引入正则化项来对参数进行过多的更新。尽管现有方法采用了一般任务无关的随机梯度下降更新规则，但我们提出了一个任务吸引的优化器，可根据任务之间的相关性调整学习率。我们通过累积针对每个任务的梯度来利用参数在更新过程中采取的方向。这些基于任务的累积梯度充当了在整个流中维护和更新的知识库。我们从经验上表明，我们提出的自适应学习率不仅说明了灾难性的遗忘，而且还允许积极的向后转移。我们还表明，在具有大量任务的复杂数据集中，我们的方法比终身学习中的几种最先进的方法更好。

大规模预训练的快速开发导致基础模型可以充当各种下游任务和领域的有效提取器。在此激励的情况下，我们研究了预训练的视觉模型的功效，作为下游持续学习（CL）场景的基础。我们的目标是双重的。首先，我们想了解RAW-DATA空间中CL和预训练编码器的潜在空间之间CL之间的计算准确性权衡。其次，我们研究编码器的特征，训练算法和数据以及所得的潜在空间如何影响CL性能。为此，我们将各种预训练的模型在大规模基准测试方案中的功效与在潜在和原始数据空间中应用的香草重播设置的功效。值得注意的是，这项研究表明了转移，遗忘，任务相似性和学习如何取决于输入数据特征，而不一定取决于CL算法。首先，我们表明，在某些情况下，通过可忽略的计算中的非参数分类器可以很容易地实现合理的CL性能。然后，我们展示模型如何在更广泛的数据上进行预训练，从而为各种重播大小提供更好的性能。我们以这些表示形式的代表性相似性和传递属性来解释这一点。最后，与训练域相比，我们显示了自我监督预训练对下游域的有效性。我们指出并验证了几个研究方向，这些方向可以进一步提高潜在CL的功效，包括表示结合。本研究中使用的各种数据集可以用作进一步CL研究的计算效率游乐场。该代码库可在https://github.com/oleksost/latent_cl下获得。