litlas
Log inCreate account

Editable Neural Networks

Authors: Anton Sinitsin, Vsevolod Plokhotnyuk, Dmitry Pyrkin, С. Б. Попов, Artem Babenko - International Conference on Learning Representations, ICLR 2020 cited by 223

These days deep neural networks are ubiquitously used in a wide range of tasks, from image classification and machine translation to face identification and self-driving cars. In many applications, a single model error can lead to devastating financial, reputational and even life-threatening consequences. Therefore, it is crucially important to correct model mistakes quickly as they appear. In this work, we investigate the problem of neural network editing $-$ how one can efficiently patch a mistake of the model on a particular sample, without influencing the model behavior on other samples. Namely, we propose Editable Training, a model-agnostic training technique that encourages fast editing of the trained model. We empirically demonstrate the effectiveness of this method on large-scale image classification and machine translation tasks.

Connectionist models ofrecognition memory…Connectionist models of recognition memory: Constraints imposed by learning and forgetting functions.Catastrophic Forgetting,Rehearsal and…Catastrophic Forgetting, Rehearsal and PseudorehearsalVisualizing Data usingt-SNEVisualizing Data using t-SNEImageNet: A large-scalehierarchical image…ImageNet: A large-scale hierarchical image databaseADADELTA: An AdaptiveLearning Rate MethodADADELTA: An Adaptive Learning Rate MethodExplaining andHarnessing Adversarial…Explaining and Harnessing Adversarial ExamplesDistilling the Knowledgein a Neural NetworkDistilling the Knowledge in a Neural NetworkDeep Residual Learningfor Image RecognitionDeep Residual Learning for Image RecognitionRethinking the InceptionArchitecture for…Rethinking the Inception Architecture for Computer VisionMastering the game of Gowith deep neural…Mastering the game of Go with deep neural networks and tree searchModel-AgnosticMeta-Learning for Fast…Model-Agnostic Meta-Learning for Fast Adaptation of Deep NetworksOvercoming catastrophicforgetting in neural…Overcoming catastrophic forgetting in neural networksEditing FactualKnowledge in Language…Editing Factual Knowledge in Language ModelsDo Language Models HaveBeliefs? Methods for…Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model BeliefsOn the Opportunities andRisks of Foundation…On the Opportunities and Risks of Foundation ModelsMeta-Learning in NeuralNetworks: A SurveyMeta-Learning in Neural Networks: A SurveyMind the Gap: AssessingTemporal Generalization…Mind the Gap: Assessing Temporal Generalization in Neural Language ModelsPatching open-vocabularymodels by interpolating…Patching open-vocabulary models by interpolating weightsTime-Aware LanguageModels as Temporal…Time-Aware Language Models as Temporal Knowledge BasesLearn From Model BeyondFine-Tuning: A SurveyLearn From Model Beyond Fine-Tuning: A SurveyStruEdit: StructuredOutputs Enable the Fast…StruEdit: Structured Outputs Enable the Fast and Accurate Knowledge Editing for Large Language ModelsAdaptive Token Biaser:Knowledge Editing via…Adaptive Token Biaser: Knowledge Editing via Biasing Key EntitiesPMET: Precise ModelEditing in a TransformerPMET: Precise Model Editing in a TransformerA Survey on KnowledgeEditing of Neural…A Survey on Knowledge Editing of Neural NetworksEditable Neural NetworksEditable Neural NetworksEarlier referencesFocus paperCiting papersOlderNewer

Click a node to pin it, click the empty canvas to go back to this paper, or hover to preview. Open a node’s page from its title.