10 Research Papers Accepted to CVPR 2023

Research from the department has been accepted to the 2023 Computer Vision and Pattern Recognition (CVPR) Conference. The annual event explores machine learning, artificial intelligence, and computer vision research and its applications. 

CoWs on Pasture: Baselines and Benchmarks for Language-Driven Zero-Shot Object Navigation
Samir Yitzhak Gadre Columbia University, Mitchell Wortsman University of Washington, Gabriel Ilharco University of Washington, Ludwig Schmidt University of Washington, Shuran Song Columbia University

For robots to be generally useful, they must be able to find arbitrary objects described by people (i.e., be language-driven) even without expensive navigation training on in-domain data (i.e., perform zero-shot inference). We explore these capabilities in a unified setting: language-driven zero-shot object navigation (L-ZSON). Inspired by the recent success of open-vocabulary models for image classification, we investigate a straightforward framework, CLIP on Wheels (CoW), to adapt open-vocabulary models to this task without fine-tuning. To better evaluate L-ZSON, we introduce the Pasture benchmark, which considers finding uncommon objects, objects described by spatial and appearance attributes, and hidden objects described relative to visible objects. We conduct an in-depth empirical study by directly deploying 21 CoW baselines across Habitat, RoboTHOR, and Pasture. In total, we evaluate over 90k navigation episodes and find that (1) CoW baselines often struggle to leverage language descriptions, but are proficient at finding uncommon objects. (2) A simple CoW, with CLIP-based object localization and classical exploration — and no additional training — matches the navigation efficiency of a state-of-the-art ZSON method trained for 500M steps on Habitat MP3D data. This same CoW provides a 15.6 percentage point improvement in success over a state-of-the-art RoboTHOR ZSON model.


Towards Fast Adaptation of Pretrained Contrastive Models for Multi-Channel Video-Language Retrieval 
Xudong Lin Columbia University, Simran Tiwari Columbia University, Shiyuan Huang Columbia University, Manling Li UIUC, Mike Zheng Shou National University of Singapore, Heng Ji UIUC, Shih-Fu Chang Columbia University

Multi-channel video-language retrieval require models to understand information from different channels (e.g. video+question, video+speech) to correctly link a video with a textual response or query. Fortunately, contrastive multimodal models are shown to be highly effective at aligning entities in images/videos and text, e.g., CLIP; text contrastive models are extensively studied recently for their strong ability of producing discriminative sentence embeddings, e.g., SimCSE. However, there is not a clear way to quickly adapt these two lines to multi-channel video-language retrieval with limited data and resources. In this paper, we identify a principled model design space with two axes: how to represent videos and how to fuse video and text information. Based on categorization of recent methods, we investigate the options of representing videos using continuous feature vectors or discrete text tokens; for the fusion method, we explore the use of a multimodal transformer or a pretrained contrastive text model. We extensively evaluate the four combinations on five video-language datasets. We surprisingly find that discrete text tokens coupled with a pretrained contrastive text model yields the best performance, which can even outperform state-of-the-art on the iVQA and How2QA datasets without additional training on millions of video-text data. Further analysis shows that this is because representing videos as text tokens captures the key visual information and text tokens are naturally aligned with text models that are strong retrievers after the contrastive pretraining process. All the empirical analysis establishes a solid foundation for future research on affordable and upgradable multimodal intelligence.


DiGeo: Discriminative Geometry-Aware Learning for Generalized Few-Shot Object Detection 
Jiawei Ma Columbia University, Yulei Niu Columbia University, Jincheng Xu Columbia University, Shiyuan Huang Columbia University, Guangxing Han Columbia University, Shih-Fu Chang Columbia University

Generalized few-shot object detection aims to achieve precise detection on both base classes with abundant annotations and novel classes with limited training data. Existing approaches enhance few-shot generalization with the sacrifice of base-class performance, or maintain high precision in base-class detection with limited improvement in novel-class adaptation. In this paper, we point out the reason is insufficient Discriminative feature learning for all of the classes. As such, we propose a new training framework, DiGeo, to learn Geometry-aware features of inter-class separation and intra-class compactness. To guide the separation of feature clusters, we derive an offline simplex equiangular tight frame (ETF) classifier whose weights serve as class centers and are maximally and equally separated. To tighten the cluster for each class, we include adaptive class-specific margins into the classification loss and encourage the features close to the class centers. Experimental studies on two few-shot benchmark datasets (VOC, COCO) and one long-tail dataset (LVIS) demonstrate that, with a single model, our method can effectively improve generalization on novel classes without hurting the detection of base classes.


Supervised Masked Knowledge Distillation for Few-Shot Transformers
Han Lin Columbia University, Guangxing Han Columbia University, Jiawei Ma Columbia University, Shiyuan Huang Columbia University, Xudong Lin Columbia University, Shih-Fu Chang Columbia University

Vision Transformers (ViTs) emerge to achieve impressive performance on many data-abundant computer vision tasks by capturing long-range dependencies among local features. However, under few-shot learning (FSL) settings on small datasets with only a few labeled data, ViT tends to overfit and suffers from severe performance degradation due to its absence of CNN-alike inductive bias. Previous works in FSL avoid such problem either through the help of self-supervised auxiliary losses, or through the dextile uses of label information under supervised settings. But the gap between self-supervised and supervised few-shot Transformers is still unfilled. Inspired by recent advances in self-supervised knowledge distillation and masked image modeling (MIM), we propose a novel Supervised Masked Knowledge Distillation model (SMKD) for few-shot Transformers which incorporates label information into self-distillation frameworks. Compared with previous self-supervised methods, we allow intra-class knowledge distillation on both class and patch tokens, and introduce the challenging task of masked patch tokens reconstruction across intra-class images. Experimental results on four few-shot classification benchmark datasets show that our method with simple design outperforms previous methods by a large margin and achieves a new start-of-the-art. Detailed ablation studies confirm the effectiveness of each component of our model. Code for this paper is available here: this https URL.


FLEX: Full-Body Grasping Without Full-Body Grasps
Purva Tendulkar Columbia University, Dídac Surís Columbia University, Carl Vondrick Columbia University

Synthesizing 3D human avatars interacting realistically with a scene is an important problem with applications in AR/VR, video games and robotics. Towards this goal, we address the task of generating a virtual human — hands and full body — grasping everyday objects. Existing methods approach this problem by collecting a 3D dataset of humans interacting with objects and training on this data. However, 1) these methods do not generalize to different object positions and orientations, or to the presence of furniture in the scene, and 2) the diversity of their generated full-body poses is very limited. In this work, we address all the above challenges to generate realistic, diverse full-body grasps in everyday scenes without requiring any 3D full-body grasping data. Our key insight is to leverage the existence of both full-body pose and hand grasping priors, composing them using 3D geometrical constraints to obtain full-body grasps. We empirically validate that these constraints can generate a variety of feasible human grasps that are superior to baselines both quantitatively and qualitatively. See our webpage for more details: this https URL.


Humans As Light Bulbs: 3D Human Reconstruction From Thermal Reflection
Ruoshi Liu Columbia University, Carl Vondrick Columbia University

The relatively hot temperature of the human body causes people to turn into long-wave infrared light sources. Since this emitted light has a larger wavelength than visible light, many surfaces in typical scenes act as infrared mirrors with strong specular reflections. We exploit the thermal reflections of a person onto objects in order to locate their position and reconstruct their pose, even if they are not visible to a normal camera. We propose an analysis-by-synthesis framework that jointly models the objects, people, and their thermal reflections, which combines generative models with differentiable rendering of reflections. Quantitative and qualitative experiments show our approach works in highly challenging cases, such as with curved mirrors or when the person is completely unseen by a normal camera.


Tracking Through Containers and Occluders in the Wild
Basile Van Hoorick Columbia University, Pavel Tokmakov Toyota Research Institute, Simon Stent Woven Planet, Jie Li Toyota Research Institute, Carl Vondrick Columbia University

Tracking objects with persistence in cluttered and dynamic environments remains a difficult challenge for computer vision systems. In this paper, we introduce TCOW, a new benchmark and model for visual tracking through heavy occlusion and containment. We set up a task where the goal is to, given a video sequence, segment both the projected extent of the target object, as well as the surrounding container or occluder whenever one exists. To study this task, we create a mixture of synthetic and annotated real datasets to support both supervised learning and structured evaluation of model performance under various forms of task variation, such as moving or nested containment. We evaluate two recent transformer-based video models and find that while they can be surprisingly capable of tracking targets under certain settings of task variation, there remains a considerable performance gap before we can claim a tracking model to have acquired a true notion of object permanence.


Doubly Right Object Recognition: A Why Prompt for Visual Rationales
Chengzhi Mao Columbia University, Revant Teotia Columbia University, Amrutha Sundar Columbia University, Sachit Menon Columbia University, Junfeng Yang Columbia University, Xin Wang Microsoft Research, Carl Vondrick Columbia University

Many visual recognition models are evaluated only on their classification accuracy, a metric for which they obtain strong performance. In this paper, we investigate whether computer vision models can also provide correct rationales for their predictions. We propose a “doubly right” object recognition benchmark, where the metric requires the model to simultaneously produce both the right labels as well as the right rationales. We find that state-of-the-art visual models, such as CLIP, often provide incorrect rationales for their categorical predictions. However, by transferring the rationales from language models into visual representations through a tailored dataset, we show that we can learn a “why prompt,” which adapts large visual representations to produce correct rationales. Visualizations and empirical experiments show that our prompts significantly improve performance on doubly right object recognition, in addition to zero-shot transfer to unseen tasks and datasets.


What You Can Reconstruct From a Shadow
Ruoshi Liu Columbia University, Sachit Menon Columbia University, Chengzhi Mao Columbia University, Dennis Park Toyota Research Institute, Simon Stent Woven Planet, Carl Vondrick Columbia University

3D reconstruction is a fundamental problem in computer vision, and the task is especially challenging when the object to reconstruct is partially or fully occluded. We introduce a method that uses the shadows cast by an unobserved object in order to infer the possible 3D volumes under occlusion. We create a differentiable image formation model that allows us to jointly infer the 3D shape of an object, its pose, and the position of a light source. Since the approach is end-to-end differentiable, we are able to integrate learned priors of object geometry in order to generate realistic 3D shapes of different object categories. Experiments and visualizations show that the method is able to generate multiple possible solutions that are consistent with the observation of the shadow. Our approach works even when the position of the light source and object pose are both unknown. Our approach is also robust to real-world images where ground-truth shadow mask is unknown.


CLIP-Sculptor: Zero-Shot Generation of High-Fidelity and Diverse Shapes From Natural Language
Aditya Sanghi Autodesk Research, Rao Fu Brown University, Vivian Liu Columbia University, Karl D.D. Willis Autodesk Research, Hooman Shayani Autodesk Research, Amir H. Khasahmadi Autodesk Research, Srinath Sridhar Brown University, Daniel Ritchie Brown University

Recent works have demonstrated that natural language can be used to generate and edit 3D shapes. However, these methods generate shapes with limited fidelity and diversity. We introduce CLIP-Sculptor, a method to address these constraints by producing high-fidelity and diverse 3D shapes without the need for (text, shape) pairs during training. CLIP-Sculptor achieves this in a multi-resolution approach that first generates in a low-dimensional latent space and then upscales to a higher resolution for improved shape fidelity. For improved shape diversity, we use a discrete latent space which is modeled using a transformer conditioned on CLIP’s image-text embedding space. We also present a novel variant of classifier-free guidance, which improves the accuracy-diversity trade-off. Finally, we perform extensive experiments demonstrating that CLIP-Sculptor outperforms state-of-the-art baselines.

7 Papers Accepted to ICLR 2023

Research papers from the department were accepted to the 11th International Conference on Learning Representations (ICLR 2023). ICLR is the premier conference on deep learning where researchers gather to discuss their work in the fields of artificial intelligence, statistics, and data science. 

Notable, top 5%

Visual Classification via Description from Large Language Models
Sachit Menon Columbia University, Carl Vondrick Columbia University

Keywords: vision-language models, CLIP, prompting, GPT-3, large language models, zero-shot recognition, multimodal

TL;DR: We enhance zero-shot recognition with vision-language models by comparing to category descriptors from GPT-3, enabling better performance in an interpretable setting that also allows for the incorporation of new concepts and bias mitigation.

Vision-language models such as CLIP have shown promising performance on a variety of recognition tasks using the standard zero-shot classification procedure — computing similarity between the query image and the embedded words for each category. By only using the category name, they neglect to make use of the rich context of additional information that language affords. The procedure gives no intermediate understanding of why a category is chosen and furthermore provides no mechanism for adjusting the criteria used towards this decision. We present an alternative framework for classification with VLMs, which we call classification by description. We ask VLMs to check for descriptive features rather than broad categories: to find a tiger, look for its stripes; its claws; and more. By basing decisions on these descriptors, we can provide additional cues that encourage using the features we want to be used. In the process, we can get a clear idea of what the model “thinks” it is seeing to make its decision; it gains some level of inherent explainability. We query large language models (e.g., GPT-3) for these descriptors to obtain them in a scalable way. Extensive experiments show our framework has numerous advantages past interpretability. We show improvements in accuracy on ImageNet across distribution shifts; demonstrate the ability to adapt VLMs to recognize concepts unseen during training; and illustrate how descriptors can be edited to effectively mitigate bias compared to the baseline.


Notable, top 25%

CROM: Continuous Reduced-Order Modeling of PDEs Using Implicit Neural Representations
Peter Yichen Chen Columbia University, Jinxu Xiang Columbia University, Dong Heon Cho Columbia University, Yue Chang University of Toronto, G A Pershing Columbia University, Henrique Teles Maia Columbia University, Maurizio M Chiaramonte Meta Reality Labs Research, Kevin Thomas Carlberg Meta Reality Labs Research, Eitan Grinspun University of Toronto

Keywords: PDE, implicit neural representation, neural field, latent space traversal, reduced-order modeling, numerical methods

TL;DR: We accelerate PDE solvers via rapid latent space traversal of continuous vector fields leveraging implicit neural representations.

The long runtime of high-fidelity partial differential equation (PDE) solvers makes them unsuitable for time-critical applications. We propose to accelerate PDE solvers using reduced-order modeling (ROM). Whereas prior ROM approaches reduce the dimensionality of discretized vector fields, our continuous reduced-order modeling (CROM) approach builds a low-dimensional embedding of the continuous vector fields themselves, not their discretization. We represent this reduced manifold using continuously differentiable neural fields, which may train on any and all available numerical solutions of the continuous system, even when they are obtained using diverse methods or discretizations. We validate our approach on an extensive range of PDEs with training data from voxel grids, meshes, and point clouds. Compared to prior discretization-dependent ROM methods, such as linear subspace proper orthogonal decomposition (POD) and nonlinear manifold neural-network-based autoencoders, CROM features higher accuracy, lower memory consumption, dynamically adaptive resolutions, and applicability to any discretization. For equal latent space dimension, CROM exhibits 79x and 49x better accuracy, and 39x and 132x smaller memory footprint, than POD and autoencoder methods, respectively. Experiments demonstrate 109x and 89x wall-clock speedups over unreduced models on CPUs and GPUs, respectively. Videos and codes are available on the project page: https://crom-pde.github.io



Quantile Risk Control: A Flexible Framework for Bounding the Probability of High-Loss Predictions
Jake Snell Princeton University, Thomas P Zollo Columbia University, Zhun Deng Columbia University, Toniann Pitassi Columbia University, Richard Zemel Columbia University

Keywords: distribution-free uncertainty quantification

TL;DR: We propose a framework to rigorously and flexible control the quantiles of the loss distribution incurred by a predictor or set of predictors.

Rigorous guarantees about the performance of predictive algorithms are necessary in order to ensure their responsible use. Previous work has largely focused on bounding the expected loss of a predictor, but this is not sufficient in many risk-sensitive applications where the distribution of errors is important. In this work, we propose a flexible framework to produce a family of bounds on quantiles of the loss distribution incurred by a predictor. Our method takes advantage of the order statistics of the observed loss values rather than relying on the sample mean alone. We show that a quantile is an informative way of quantifying predictive performance, and that our framework applies to a variety of quantile-based metrics, each targeting important subsets of the data distribution. We analyze the theoretical properties of our proposed method and demonstrate its ability to rigorously control loss quantiles on several real-world datasets.

Causal Imitation Learning via Inverse Reinforcement Learning
Kangrui Ruan Columbia University, Junzhe Zhang Columbia University, Xuan Di Columbia University, Elias Bareinboim Columbia University

Keywords: Causal Inference, Graphical Models

TL;DR: This paper proposes novel inverse reinforcement learning methods to learn effective imitating policies from the expert’s demonstrations when unobserved confounders are present.

One of the most common ways children learn when unfamiliar with the environment is by mimicking adults. Imitation learning concerns an imitator learning to behave in an unknown environment from an expert’s demonstration; reward signals remain latent to the imitator. This paper studies imitation learning through causal lenses and extends the analysis and tools developed for behavior cloning (Zhang, Kumor, Bareinboim, 2020) to inverse reinforcement learning. First, we propose novel graphical conditions that allow the imitator to learn a policy performing as well as the expert’s behavior policy, even when the imitator and the expert’s state-action space disagree, and unobserved confounders (UCs) are present. When provided with parametric knowledge about the unknown reward function, such a policy may outperform the expert’s. Also, our method is easily extensible and allows one to leverage existing IRL algorithms even when UCs are present, including the multiplicative-weights algorithm (MWAL) (Syed & Schapire, 2008) and the generative adversarial imitation learning (GAIL) (Ho & Ermon, 2016). Finally, we validate our framework by simulations using real-world and synthetic data.

Neural Causal Models for Counterfactual Identification and Estimation
Kevin Muyuan Xia Columbia University, Yushu Pan Columbia University, Elias Bareinboim Columbia University

Keywords: causal inference, deep learning, neural models, neural causal models, causal identification, causal estimation, counterfactual

TL;DR: We solve the two problems of counterfactual identification and estimation from arbitrary surrogate experiments using a Generative Adversarial Network implementation of the Neural Causal Model.

Evaluating hypothetical statements about how the world would be had a different course of action been taken is arguably one key capability expected from modern AI systems. Counterfactual reasoning underpins discussions in fairness, the determination of blame and responsibility, credit assignment, and regret. In this paper, we study the evaluation of counterfactual statements through neural models. Specifically, we tackle two causal problems required to make such evaluations, i.e., counterfactual identification and estimation from an arbitrary combination of observational and experimental data. First, we show that neural causal models (NCMs) are expressive enough and encode the structural constraints necessary for performing counterfactual reasoning. Second, we develop an algorithm for simultaneously identifying and estimating counterfactual distributions. We show that this algorithm is sound and complete for deciding counterfactual identification in general settings. Third, considering the practical implications of these results, we introduce a new strategy for modeling NCMs using generative adversarial networks. Simulations corroborate with the proposed methodology.

Understanding Zero-shot Adversarial Robustness for Large-Scale Models
Chengzhi Mao Columbia University, Scott Geng Columbia University, Junfeng Yang Columbia University, Xin Wang Microsoft Research, Carl Vondrick Columbia University

Keywords: Adversarial Robustness, Zero-Shot Recognition

Pretrained large-scale vision-language models like CLIP have exhibited strong generalization over unseen tasks. Yet imperceptible adversarial perturbations can significantly reduce CLIP’s performance on new tasks. In this work, we identify and explore the problem of adapting large-scale models for zero-shot adversarial robustness. We first identify two key factors during model adaption–training losses and adaptation methods–that affect the model’s zero-shot adversarial robustness. We then propose a text-guided contrastive adversarial training loss, which aligns the text embeddings and the adversarial visual features with contrastive learning on a small set of training data. We apply this training loss to two adaption methods, model finetuning and visual prompt tuning. We find that visual prompt tuning is more effective in the absence of texts, while finetuning wins in the existence of text guidance. Overall, our approach significantly improves the zero-shot adversarial robustness over CLIP, seeing an average improvement of 31 points over ImageNet and 15 zero-shot datasets. We hope this work can shed light on understanding the zero-shot adversarial robustness of large-scale models.

TempCLR: Temporal Alignment Representation with Contrastive Learning
Yuncong Yang Columbia University, Jiawei Ma Columbia University, Shiyuan Huang Columbia University, Long Chen Columbia University, Xudong Lin Columbia University, Guangxing Han Columbia University, Shih-Fu Chang Columbia University

Keywords: Representation learning, Global Sequence Alignment, Zero/Few-shot Transfer

TL;DR: Global sequence matching under temporal order consistency matters in contrastive-based video-paragraph/text learning.

Video representation learning has been successful in video-text pre-training for zero-shot transfer, where each sentence is trained to be close to the paired video clips in a common feature space. For long videos, given a paragraph of description where the sentences describe different segments of the video, by matching all sentence-clip pairs, the paragraph and the full video are aligned implicitly. However, such unit-level similarity measure may ignore the global temporal context over a long time span, which inevitably limits the generalization ability. In this paper, we propose a contrastive learning framework TempCLR to compare the full video and the paragraph explicitly. As the video/paragraph is formulated as a sequence of clips/sentences, under the constraint of their temporal order, we use dynamic time warping to compute the minimum cumulative cost over sentence-clip pairs as the sequence-level distance. To explore the temporal dynamics, we break the consistency of temporal order by shuffling the video clips or sentences according to the temporal granularity. In this way, we obtain the representations for clips/sentences, which perceive the temporal information and thus facilitate the sequence alignment. In addition to pre-training on the video and paragraph, our approach can also generalize on the matching between different video instances. We evaluate our approach on video retrieval, action step localization, and few-shot action recognition, and achieve consistent performance gain over all three tasks. Detailed ablation studies are provided to justify the approach design.

6 Papers from CS Researchers Accepted to NAACL 2022

Researchers from the department presented natural language processing (NLP) papers at the 2022 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2022).

Selective Differential Privacy for Language Models
Weiyan Shi, Aiqi Cui, Evan Li, Ruoxi Jia, Zhou Yu

With the increasing applications of language models, it has become crucial to protect these models from leaking private information. Previous work has attempted to tackle this challenge by training RNN-based language models with differential privacy guarantees. However, applying classical differential privacy to language models leads to poor model performance as the underlying privacy notion is over-pessimistic and provides undifferentiated protection for all tokens in the data. Given that the private information in natural language is sparse (for example, the bulk of an email might not carry personally identifiable information), we propose a new privacy notion, selective differential privacy, to provide rigorous privacy guarantees on the sensitive portion of the data to improve model utility. To realize such a new notion, we develop a corresponding privacy mechanism, Selective-DPSGD, for RNN-based language models. Besides language modeling, we also apply the method to a more concrete application–dialog systems. Experiments on both language modeling and dialog system building show that the proposed privacy-preserving mechanism achieves better utilities while remaining safe under various privacy attacks compared to the baselines. The data and code are released at this HTTPS URL to facilitate future research.

Knowledge-Grounded Dialogue Generation with a Unified Knowledge Representation
Yu Li, Baolin Peng, Yelong Shen, Yi Mao, Lars Liden, Zhou Yu, Jianfeng Gao

Knowledge-grounded dialogue systems are challenging to build due to the lack of training data and heterogeneous knowledge sources. Existing systems perform poorly on unseen topics due to limited topics covered in the training data. In addition, heterogeneous knowledge sources make it challenging for systems to generalize to other tasks because knowledge sources in different knowledge representations require different knowledge encoders. To address these challenges, we present PLUG, a language model that homogenizes different knowledge sources to a unified knowledge representation for knowledge-grounded dialogue generation tasks. PLUG is pre-trained on a dialogue generation task conditioned on a unified essential knowledge representation. It can generalize to different downstream knowledge-grounded dialogue generation tasks with a few training examples. The empirical evaluation on two benchmarks shows that our model generalizes well across different knowledge-grounded tasks. It can achieve comparable performance with state-of-the-art methods under a fully-supervised setting and significantly outperforms other methods in zero-shot and few-shot settings.

Database Search Results Disambiguation for Task-Oriented Dialog Systems
Kun Qian, Ahmad Beirami, Satwik Kottur, Shahin Shayandeh, Paul Crook, Alborz Geramifard, Zhou Yu, Chinnadhurai Sankar

As task-oriented dialog systems are becoming increasingly popular in our lives, more realistic tasks have been proposed and explored. However, new practical challenges arise. For instance, current dialog systems cannot effectively handle multiple search results when querying a database, due to the lack of such scenarios in existing public datasets. In this paper, we propose Database Search Result (DSR) Disambiguation, a novel task that focuses on disambiguating database search results, which enhances user experience by allowing them to choose from multiple options instead of just one. To study this task, we augment the popular task-oriented dialog datasets (MultiWOZ and SGD) with turns that resolve ambiguities by (a) synthetically generating turns through a pre-defined grammar, and (b) collecting human paraphrases for a subset. We find that training on our augmented dialog data improves the model’s ability to deal with ambiguous scenarios, without sacrificing performance on unmodified turns. Furthermore, pre-fine tuning and multi-task learning help our model to improve performance on DSRdisambiguation even in the absence of indomain data, suggesting that it can be learned as a universal dialog skill. Our data and code will be made publicly available.

ErAConD: Error Annotated Conversational Dialog Dataset for Grammatical Error Correction
Xun Yuan, Sam Pham, Sam Davidson, Zhou Yu

Currently available grammatical error correction (GEC) datasets are compiled using well-formed written text, limiting the applicability of these datasets to other domains such as informal writing and dialog. In this paper, we present a novel parallel GEC dataset drawn from open-domain chatbot conversations; this dataset is, to our knowledge, the first GEC dataset targeted to a conversational setting. To demonstrate the utility of the dataset, we use our annotated data to fine-tune a state-of-the-art GEC model, resulting in a 16-point increase in model precision. This is of particular importance in a GEC model, as model precision is considered more important than recall in GEC tasks since false positives could lead to serious confusion in language learners. We also present a detailed annotation scheme which ranks errors by perceived impact on comprehensibility, making our dataset both reproducible and extensible. Experimental results show the effectiveness of our data in improving GEC model performance in conversational scenarios.

Improving Conversational Recommendation Systems’ Quality with Context-Aware Item Meta-Information
Bowen Yang, Cong Han, Yu Li, Lei Zuo, Zhou Yu

Conversational recommendation systems (CRS) engage with users by inferring user preferences from dialog history, providing accurate recommendations, and generating appropriate responses. Previous CRSs use knowledge graph (KG) based recommendation modules and integrate KG with language models for response generation. Although KG-based approaches prove effective, two issues remain to be solved. First, KG-based approaches ignore the information in the conversational context but only rely on entity relations and bag of words to recommend items. Second, it requires substantial engineering efforts to maintain KGs that model domain-specific relations, thus leading to less flexibility. In this paper, we propose a simple yet effective architecture comprising a pre-trained language model (PLM) and an item metadata encoder. The encoder learns to map item metadata to embeddings that can reflect the semantic information in the dialog context. The PLM then consumes the semantic-aligned item embeddings together with dialog context to generate high-quality recommendations and responses. Instead of modeling entity relations with KGs, our model reduces engineering complexity by directly converting each item to an embedding. Experimental results on the benchmark dataset ReDial show that our model obtains state-of-the-art results on both recommendation and response generation tasks.

Differentially private decoding in large language models
By Jimit Majmudar, Christophe Dupuy, Charith Peris, Sami Smaili, Rahul Gupta, Richard Zemel

Recent large-scale natural language processing (NLP) systems use a pre-trained Large Language Model (LLM) on massive and diverse corpora as a headstart. In practice, the pre-trained model is adapted to a wide array of tasks via fine-tuning on task-specific datasets. LLMs, while effective, have been shown to memorize instances of training data thereby potentially revealing private information processed during pre-training. The potential leakage might further propagate to the downstream tasks for which LLMs are fine-tuned. On the other hand, privacy-preserving algorithms usually involve retraining from scratch, which is prohibitively expensive for LLMs. In this work, we propose a simple, easy to interpret, and computationally lightweight perturbation mechanism to be applied to an already trained model at the decoding stage. Our perturbation mechanism is model-agnostic and can be used in conjunction with any LLM. We provide a theoretical analysis showing that the proposed mechanism is differentially private, and experimental results show a privacy-utility trade-off.

13 Research Papers Accepted to ICML 2021

Papers from CS researchers have been accepted to the 38th International Conference on Machine Learning (ICML 2021). 

Associate Professor Daniel Hsu was one of the publication chairs of the conference and Assistant Professor Elham Azizi helped organize the 2021 ICML Workshop on Computational Biology. The workshop highlighted how machine learning approaches can be tailored to making both translational and basic scientific discoveries with biological data.

Below are the abstracts and links to the accepted papers.

A Proxy Variable View of Shared Confounding
Yixin Wang Columbia University, David Blei Columbia University

Causal inference from observational data can be biased by unobserved confounders. Confounders—the variables that affect both the treatments and the outcome—induce spurious non-causal correlations between the two. Without additional conditions, unobserved confounders generally make causal quantities hard to identify. In this paper, we focus on the setting where there are many treatments with shared confounding, and we study under what conditions is causal identification possible. The key observation is that we can view subsets of treatments as proxies of the unobserved confounder and identify the intervention distributions of the rest. Moreover, while existing identification formulas for proxy variables involve solving integral equations, we show that one can circumvent the need for such solutions by directly modeling the data. Finally, we extend these results to an expanded class of causal graphs, those with other confounders and selection variables.


Unsupervised Representation Learning via Neural Activation Coding
Yookoon Park Columbia University, Sangho Lee Seoul National University, Gunhee Kim Seoul National University, David Blei Columbia University

We present neural activation coding (NAC) as a novel approach for learning deep representations from unlabeled data for downstream applications. We argue that the deep encoder should maximize its nonlinear expressivity on the data for downstream predictors to take full advantage of its representation power. To this end, NAC maximizes the mutual information between activation patterns of the encoder and the data over a noisy communication channel. We show that learning for a noise-robust activation code increases the number of distinct linear regions of ReLU encoders, hence the maximum nonlinear expressivity. More interestingly, NAC learns both continuous and discrete representations of data, which we respectively evaluate on two downstream tasks: (i) linear classification on CIFAR-10 and ImageNet-1K and (ii) nearest neighbor retrieval on CIFAR-10 and FLICKR-25K. Empirical results show that NAC attains better or comparable performance on both tasks over recent baselines including SimCLR and DistillHash. In addition, NAC pretraining provides significant benefits to the training of deep generative models. Our code is available at https://github.com/yookoon/nac.


The Logical Options Framework
Brandon Araki MIT, Xiao Li MIT, Kiran Vodrahalli Columbia University, Jonathan DeCastro Toyota Research Institute, Micah Fry MIT Lincoln Laboratory, Daniela Rus MIT CSAIL

Learning composable policies for environments with complex rules and tasks is a challenging problem. We introduce a hierarchical reinforcement learning framework called the Logical Options Framework (LOF) that learns policies that are satisfying, optimal, and composable. LOF efficiently learns policies that satisfy tasks by representing the task as an automaton and integrating it into learning and planning. We provide and prove conditions under which LOF will learn satisfying, optimal policies. And lastly, we show how LOF’s learned policies can be composed to satisfy unseen tasks with only 10-50 retraining steps on our benchmarks. We evaluate LOF on four tasks in discrete and continuous domains, including a 3D pick-and-place environment.


Estimating Identifiable Causal Effects on Markov Equivalence Class through Double Machine Learning
Yonghan Jung Columbia University, Jin Tian Columbia University, Elias Bareinboim Columbia University

General methods have been developed for estimating causal effects from observational data under causal assumptions encoded in the form of a causal graph. Most of this literature assumes that the underlying causal graph is completely specified. However, only observational data is available in most practical settings, which means that one can learn at most a Markov equivalence class (MEC) of the underlying causal graph. In this paper, we study the problem of causal estimation from a MEC represented by a partial ancestral graph (PAG), which is learnable from observational data. We develop a general estimator for any identifiable causal effects in a PAG. The result fills a gap for an end-to-end solution to causal inference from observational data to effects estimation. Specifically, we develop a complete identification algorithm that derives an influence function for any identifiable causal effects from PAGs. We then construct a double/debiased machine learning (DML) estimator that is robust to model misspecification and biases in nuisance function estimation, permitting the use of modern machine learning techniques. Simulation results corroborate with the theory.


Environment Inference for Invariant Learning 
Elliot Creager University of Toronto, Joern Jacobsen Apple Inc., Richard Zemel Columbia University

Learning models that gracefully handle distribution shifts is central to research on domain generalization, robust optimization, and fairness. A promising formulation is domain-invariant learning, which identifies the key issue of learning which features are domain-specific versus domain-invariant. An important assumption in this area is that the training examples are partitioned into domains'' orenvironments”. Our focus is on the more common setting where such partitions are not provided. We propose EIIL, a general framework for domain-invariant learning that incorporates Environment Inference to directly infer partitions that are maximally informative for downstream Invariant Learning. We show that EIIL outperforms invariant learning methods on the CMNIST benchmark without using environment labels, and significantly outperforms ERM on worst-group performance in the Waterbirds dataset. Finally, we establish connections between EIIL and algorithmic fairness, which enables EIIL to improve accuracy and calibration in a fair prediction problem.


SketchEmbedNet: Learning Novel Concepts by Imitating Drawings
Alex Wang University of Toronto, Mengye Ren University of Toronto, Richard Zemel Columbia University

Sketch drawings capture the salient information of visual concepts. Previous work has shown that neural networks are capable of producing sketches of natural objects drawn from a small number of classes. While earlier approaches focus on generation quality or retrieval, we explore properties of image representations learned by training a model to produce sketches of images. We show that this generative, class-agnostic model produces informative embeddings of images from novel examples, classes, and even novel datasets in a few-shot setting. Additionally, we find that these learned representations exhibit interesting structure and compositionality.


Universal Template for Few-Shot Dataset Generalization
Eleni Triantafillou University of Toronto, Hugo Larochelle Google Brain, Richard Zemel Columbia University, Vincent Dumoulin Google

Few-shot dataset generalization is a challenging variant of the well-studied few-shot classification problem where a diverse training set of several datasets is given, for the purpose of training an adaptable model that can then learn classes from \emph{new datasets} using only a few examples. To this end, we propose to utilize the diverse training set to construct a \emph{universal template}: a partial model that can define a wide array of dataset-specialized models, by plugging in appropriate components. For each new few-shot classification problem, our approach therefore only requires inferring a small number of parameters to insert into the universal template. We design a separate network that produces an initialization of those parameters for each given task, and we then fine-tune its proposed initialization via a few steps of gradient descent. Our approach is more parameter-efficient, scalable and adaptable compared to previous methods, and achieves the state-of-the-art on the challenging Meta-Dataset benchmark.


On Monotonic Linear Interpolation of Neural Network Parameters
James Lucas University of Toronto, Juhan Bae University of Toronto, Michael Zhang University of Toronto, Stanislav Fort Google AI, Richard Zemel Columbia University, Roger Grosse University of Toronto

Linear interpolation between initial neural network parameters and converged parameters after training with stochastic gradient descent (SGD) typically leads to a monotonic decrease in the training objective. This Monotonic Linear Interpolation (MLI) property, first observed by Goodfellow et al. 2014, persists in spite of the non-convex objectives and highly non-linear training dynamics of neural networks. Extending this work, we evaluate several hypotheses for this property that, to our knowledge, have not yet been explored. Using tools from differential geometry, we draw connections between the interpolated paths in function space and the monotonicity of the network — providing sufficient conditions for the MLI property under mean squared error. While the MLI property holds under various settings (e.g., network architectures and learning problems), we show in practice that networks violating the MLI property can be produced systematically, by encouraging the weights to move far from initialization. The MLI property raises important questions about the loss landscape geometry of neural networks and highlights the need to further study their global properties.


A Computational Framework For Slang Generation
Zhewei Sun University of Toronto, Richard Zemel Columbia University, Yang Xu University of Toronto

Slang is a common type of informal language, but its flexible nature and paucity of data resources present challenges for existing natural language systems. We take an initial step toward machine generation of slang by developing a framework that models the speaker’s word choice in slang context. Our framework encodes novel slang meaning by relating the conventional and slang senses of a word while incorporating syntactic and contextual knowledge in slang usage. We construct the framework using a combination of probabilistic inference and neural contrastive learning. We perform rigorous evaluations on three slang dictionaries and show that our approach not only outperforms state-of-the-art language models, but also better predicts the historical emergence of slang word usages from 1960s to 2000s. We interpret the proposed models and find that the contrastively learned semantic space is sensitive to the similarities between slang and conventional senses of words. Our work creates opportunities for the automated generation and interpretation of informal language.


Wandering Within A World: Online Contextualized Few-Shot Learning
Mengye Ren University of Toronto, Michael Iuzzolino Google Research, Michael Mozer Google Research, Richard Zemel Columbia University

We aim to bridge the gap between typical human and machine-learning environments by extending the standard framework of few-shot learning to an online, continual setting. In this setting, episodes do not have separate training and testing phases, and instead models are evaluated online while learning novel classes. As in the real world, where the presence of spatiotemporal context helps us retrieve learned skills in the past, our online few-shot learning setting also features an underlying context that changes throughout time. Object classes are correlated within a context and inferring the correct context can lead to better performance. Building upon this setting, we propose a new few-shot learning dataset based on large scale indoor imagery that mimics the visual experience of an agent wandering within a world. Furthermore, we convert popular few-shot learning approaches into online versions and we also propose a new contextual prototypical memory model that can make use of spatiotemporal contextual information from the recent past.


Bayesian Few-Shot Classification With One-Vs-Each Polya-Gamma Augmented Gaussian Processes
Jake Snell University of Toronto, Richard Zemel Columbia University

Few-shot classification (FSC), the task of adapting a classifier to unseen classes given a small labeled dataset, is an important step on the path toward human-like machine learning. Bayesian methods are well-suited to tackling the fundamental issue of overfitting in the few-shot scenario because they allow practitioners to specify prior beliefs and update those beliefs in light of observed data. Contemporary approaches to Bayesian few-shot classification maintain a posterior distribution over model parameters, which is slow and requires storage that scales with model size. Instead, we propose a Gaussian process classifier based on a novel combination of Pólya-Gamma augmentation and the one-vs-each softmax approximation that allows us to efficiently marginalize over functions rather than model parameters. We demonstrate improved accuracy and uncertainty quantification on both standard few-shot classification benchmarks and few-shot domain transfer tasks.


Theoretical Bounds On Estimation Error For Meta-Learning
James Lucas University of Toronto, Mengye Ren University of Toronto, Irene Kameni African Master for Mathematical Sciences, Toni Pitassi Columbia University, Richard Zemel Columbia University

Machine learning models have traditionally been developed under the assumption that the training and test distributions match exactly. However, recent success in few-shot learning and related problems are encouraging signs that these models can be adapted to more realistic settings where train and test distributions differ. Unfortunately, there is severely limited theoretical support for these algorithms and little is known about the difficulty of these problems. In this work, we provide novel information-theoretic lower-bounds on minimax rates of convergence for algorithms that are trained on data from multiple sources and tested on novel data. Our bounds depend intuitively on the information shared between sources of data, and characterize the difficulty of learning in this setting for arbitrary algorithms. We demonstrate these bounds on a hierarchical Bayesian model of meta-learning, computing both upper and lower bounds on parameter estimation via maximum-a-posteriori inference.


A PAC-Bayesian Approach To Generalization Bounds For Graph Neural Networks
Renjie Liao University of Toronto, Raquel Urtasun University of Toronto, Richard Zemel Columbia University

In this paper, we derive generalization bounds for the two primary classes of graph neural networks (GNNs), namely graph convolutional networks (GCNs) and message passing GNNs (MPGNNs), via a PAC-Bayesian approach. Our result reveals that the maximum node degree and spectral norm of the weights govern the generalization bounds of both models. We also show that our bound for GCNs is a natural generalization of the results developed in arXiv:1707.09564v2 [cs.LG] for fully-connected and convolutional neural networks. For message passing GNNs, our PAC-Bayes bound improves over the Rademacher complexity based bound in arXiv:2002.06157v1 [cs.LG], showing a tighter dependency on the maximum node degree and the maximum hidden dimension. The key ingredients of our proofs are a perturbation analysis of GNNs and the generalization of PAC-Bayes analysis to non-homogeneous GNNs. We perform an empirical study on several real-world graph datasets and verify that our PAC-Bayes bound is tighter than others.

6 Papers From the Department Accepted to the EACL 2021

Six papers from CS researchers were accepted to the 16th conference of the European Chapter of the Association for Computational Linguistics (EACL).  As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.

Below are brief descriptions and links to the papers. 

Event-Driven News Stream Clustering using Entity-Aware Contextual Embeddings
Kailash Karthik Saravanakumar Columbia University, Miguel Ballesteros Amazon AI, Muthu Kumar Chandrasekaran Amazon AI, Kathleen McKeown Columbia University & Amazon AI

This paper presents a new clustering paradigm for news streams, where clusters have a one-to-one correspondence with real-world events (for example, the Suez canal blockage). An important aspect of this problem is that the number of clusters is unknown and varies with time (new events occur and old events cease to be of relevance). The proposed paradigm follows a pipeline approach – where representations are built for each new article, comparisons are made with existing clusters to pick the most compatible one, and finally, a clustering decision is produced.

A surprising observation from this work is that contextual embeddings (from models like BERT), in contrast to their overwhelming success in many NLP problems, achieve sub-par performance by themselves on this clustering problem. However, when combined with other representations (like TF-IDF and timestamps) and fine-tuned with task-specific augmentations, they achieve new state-of-the-art performance. Another interesting observation is that the widely reported B-Cubed metrics are biased towards large clusters and hence don’t capture cluster fragmentation on smaller clusters as well. Since clusters corresponding to emerging events are small and errors made on such clusters are highly undesirable, the authors suggest using an additional metric CEAF-e to evaluate models for this task.


Segmenting Subtitles for Correcting ASR Segmentation Errors 
David Wan Columbia University, Chris Kedzie Columbia University, Faisal Ladakh Columbia University, Elsbeth Turcan Columbia University, Petra Galuszkova University of Maryland, Elena Zotkina University of Maryland, Zhengping Jiang Columbia University, Peter Bell University of Edinburgh, and Kathleen McKeown Columbia University

For the task of spoken language translation, the usual approach is to have a pipeline consisting of Automatic Speech Recognition (ASR) that transforms audio files into words and utterances in the original language and a Machine Translation (MT) that translate the utterances into the target language. However this setup may suffer from input-output mismatches: ASR segments utterances by acoustic information such as pauses, and thus may produce run-on sentences or sentence fragments, but MT is usually trained on proper sentences without such issues and may not perform well under such setting. This paper proposes the use of an intermediate model to segment utterances into sentences to improve performance in MT as well as other downstream tasks.

One crucial problem for developing such models is the lack of suitable training data for segmentation, especially when the languages involved are low-resourced. To this end, this paper also proposes a way to use subtitles dataset as proxy speech data as well as creating synthetic acoustic utterances that mimic common ASR errors for the model to learn to fix. Using a simple neural tagging model, the authors of this paper show improvement over the baseline ASR segmentation on MT for Lithuanian, Bulgarian, and Farisi. A surprising finding is that the segmentation model most improves the translation quality of more syntactically complex segments.


“Talk to me with left, right, and angles”: Lexical entrainment in spoken Hebrew dialogue
Andreas Weise CUNY Graduate Center, Vered Silber-Varod The Open University of Israel, Anat Lerner The Open University of Israel, Julia Hirschberg Columbia University, and Rivka Levitan Columbia University

It has been well-documented for several languages that human interlocutors tend to adapt their linguistic productions to become more similar to each other. This behavior, known as entrainment, affects lexical choice as well, both with regard to specific words, such as referring expressions, and overall style.

Lexical entrainment is the behavior that causes the words that speakers use in a conversation to become more similar over time. Entrainment more broadly is a human behavior causing interlocutors to adapt to each other to become more similar. Its effects are measurable but entrainment itself is not a measure.

This paper offers the first investigation of such lexical entrainment in Hebrew.

The analysis of Hebrew speakers interacting in a Map Task, a popular experimental setup, provides rich evidence of lexical entrainment. No clear pattern of differences is found between speaker pairs by the combination of their genders, nor between speakers by their individual gender. However, speakers in a position of less power are found to entrain more than those with greater power, which matches theoretical accounts.

Overall, the results mostly accord with those for American English. There is, however, a surprising lack of entrainment on a list of hedge words that were previously found to be highly entrained in English. This might be due to cultural differences between American and Israeli speakers that render adoption of a more tentative style less appropriate in the Hebrew context.


Entity-level Factual Consistency of Abstractive Text Summarization
Feng Nan Amazon Web Services, Ramesh Nallapati Amazon Web Services, Zhiguo Wang Amazon Web Services, Cicero Nogueira dos Santos Amazon Web Services, Henghui Zhu Amazon Web Services, Dejiao Zhang Amazon Web Services, Kathleen McKeown Amazon Web Services & Columbia University, Bing Xiang Amazon Web Services

A key challenge for abstractive summarization is ensuring factual consistency of the generated summary with respect to the original document. For example, state-of-the-art models trained on existing datasets exhibit entity hallucination, generating names of entities that are not present in the source document.

The paper proposes a set of new metrics to quantify the entity-level factual consistency of generated summaries and shows that the entity hallucination problem can be alleviated by simply filtering the training data. In addition, the paper introduces a summary-worthy entity classification task to the training process as well as a joint entity and summary generation approach, which yields further improvements in entity-level metrics.


“Laughing at you or with you”: The Role of Sarcasm in Shaping the Disagreement Space 
Debanjan Ghosh Educational Testing Service, Ritvik Shrivastava MindMeld, Cisco Systems & Columbia University, and Smaranda Muresan Columbia University

Detecting arguments in online interactions is useful to understand how conflicts arise and get resolved. Users often use figurative language, such as sarcasm, either as persuasive devices or to attack the opponent by an ad hominem argument. To further our understanding of the role of sarcasm in shaping the disagreement space, the paper presents a thorough experimental setup using a corpus annotated with both argumentative moves (agree/disagree) and sarcasm. The research exploits joint modeling in terms of (a) applying discrete features that are useful in detecting sarcasm to the task of argumentative relation classification (agree/disagree/none), and (b) multitask learning for argumentative relation classification and sarcasm detection using deep learning architectures (e.g., dual Long ShortTerm Memory (LSTM) with hierarchical attention and Transformer-based architectures). The paper shows that modeling sarcasm improves the argumentative relation classification task (agree/disagree/none) in all setups.


A Unified Feature Representation for Lexical Connotations 
Emily Allaway Columbia University and Kathleen McKeown Columbia University

Ideological attitudes and stances are often expressed through subtle meanings of words and phrases. Understanding these connotations is critical to recognize the cultural and emotional perspectives of the speaker. In this paper, the researchers use distant labeling to create a new lexical resource representing connotation aspects for nouns and adjectives. Their analysis shows that it aligns well with human judgments. Additionally, they present a method for creating lexical representations that capture connotations within the embedding space and show that using the embeddings provides a statistically significant improvement on the task of stance detection when data is limited.