Publications

Questions Are All You Need to Train a Dense Passage Retriever

Devendra Singh Sachan

Mike Lewis

Dani Yogatama

Luke Zettlemoyer

Manzil Zaheer

We introduce ART, a new corpus-level autoencoding approach for training dense retrieval models that does not require any labeled training da… (see more)ta. Dense retrieval is a central challenge for open-domain tasks, such as Open QA, where state-of-the-art methods typically require large supervised datasets with custom hard-negative mining and denoising of positive examples. ART, in contrast, only requires access to unpaired inputs and outputs (e.g. questions and potential answer documents). It uses a new document-retrieval autoencoding scheme, where (1) an input question is used to retrieve a set of evidence documents, and (2) the documents are then used to compute the probability of reconstructing the original question. Training for retrieval based on question reconstruction enables effective unsupervised learning of both document and question encoders, which can be later incorporated into complete Open QA systems without any further finetuning. Extensive experiments demonstrate that ART obtains state-of-the-art results on multiple QA retrieval benchmarks with only generic initialization from a pre-trained language model, removing the need for labeled data and task-specific losses.

2022-12-31

Transactions of the Association for Computational Linguistics (published)

doi.org

arxiv.org

A rapid review for developing a co-design framework for a pediatric surgical communication application

Michelle Cwintal

Hamed Ranjbar

Parsa Bandamiri

Elena Guadagno

Esli Osmanlliu

Dan Poenaru

2022-12-31

Journal of Pediatric Surgery (published)

doi.org

Real-Time Simultaneous Localization and Mapping with LiDAR intensity

Wenqiang Du

Giovanni Beltrame

We propose a novel real-time LiDAR intensity image-based simultaneous localization and mapping method , which addresses the geometry degener… (see more)acy problem in unstructured environments. Traditional LiDAR-based front-end odometry mostly relies on geometric features such as points, lines and planes. A lack of these features in the environment can lead to the failure of the entire odometry system. To avoid this problem, we extract feature points from the LiDAR-generated point cloud that match features identified in LiDAR intensity images. We then use the extracted feature points to perform scan registration and estimate the robot ego-movement. For the back-end, we jointly optimize the distance between the corresponding feature points, and the point to plane distance for planes identified in the map. In addition, we use the features extracted from intensity images to detect loop closure candidates from previous scans and perform pose graph optimization. Our experiments show that our method can run in real time with high accuracy and works well with illumination changes, low-texture, and unstructured environments.

2022-12-31

IEEE International Conference on Robotics and Automation (published)

doi.org

arxiv.org

Recall as a Measure of Ranking Robustness

Fernando Diaz

Bhaskar Mitra

2022-12-31

arXiv.org (preprint)

doi.org

Relationship between prediction accuracy and feature importance reliability: An empirical and theoretical study

Jianzhong Chen

Leon Qi Rong Ooi

Trevor Wei Kiat Tan

Shaoshi Zhang

Jingwei Li

Christopher L. Asplund

Simon B. Eickhoff

Danilo Bzdok

Avram J. Holmes

B.T. Thomas Yeo

There is significant interest in using neuroimaging data to predict behavior. The predictive models are often interpreted by the computation… (see more) of feature importance, which quantifies the predictive relevance of an imaging feature. Tian and Zalesky (2021) suggest that feature importance estimates exhibit low split-half reliability, as well as a trade-off between prediction accuracy and feature importance reliability across parcellation resolutions. However, it is unclear whether the trade-off between prediction accuracy and feature importance reliability is universal. Here, we demonstrate that, with a sufficient sample size, feature importance (operationalized as Haufe-transformed weights) can achieve fair to excellent split-half reliability. With a sample size of 2600 participants, Haufe-transformed weights achieve average intra-class correlation coefficients of 0.75, 0.57 and 0.53 for cognitive, personality and mental health measures respectively. Haufe-transformed weights are much more reliable than original regression weights and univariate FC-behavior correlations. Original regression weights are not reliable even with 2600 participants. Intriguingly, feature importance reliability is strongly positively correlated with prediction accuracy across phenotypes. Within a particular behavioral domain, there is no clear relationship between prediction performance and feature importance reliability across regression models. Furthermore, we show mathematically that feature importance reliability is necessary, but not sufficient, for low feature importance error. In the case of linear models, lower feature importance error is mathematically related to lower prediction error. Therefore, higher feature importance reliability might yield lower feature importance error and higher prediction accuracy. Finally, we discuss how our theoretical results relate with the reliability of imaging features and behavioral measures. Overall, the current study provides empirical and theoretical insights into the relationship between prediction accuracy and feature importance reliability.

2022-12-31

NeuroImage (published)

doi.org

A reproducible benchmark of resting-state fMRI denoising strategies using fMRIPrep and Nilearn

Hao-Ting Wang

Steven L. Meisler

Hanad Sharmarke

Natasha Clarke

Nicolas Gensollen

Christopher J Markiewicz

Fraçois Paugam

Bertrand Thirion

Lune P Bellec

Reducing contributions from non-neuronal sources is a crucial step in functional magnetic resonance imaging (fMRI) analyses. Many viable str… (see more)ategies for denoising fMRI are used in the literature, and practitioners rely on denoising benchmarks for guidance in the selection of an appropriate choice for their study. However, fMRI denoising software is an ever-evolving field, and the benchmarks can quickly become obsolete as the techniques or implementations change. In this work, we present a fully reproducible denoising benchmark featuring a range of denoising strategies and evaluation metrics, built primarily on the fMRIPrep and Nilearn software packages. We apply this reproducible benchmark to investigate the robustness of the conclusions across two different datasets and two versions of fMRIPrep. The majority of benchmark results were consistent with prior literature. Scrubbing, a technique which excludes time points with excessive motion, combined with global signal regression, is generally effective at noise removal. Scrubbing however disrupts the continuous sampling of brain images and is incompatible with some statistical analyses, e.g. auto-regressive modeling. In this case, a simple strategy using motion parameters, average activity in select brain compartments, and global signal regression should be preferred. Importantly, we found that certain denoising strategies behave inconsistently across datasets and/or versions of fMRIPrep, or had a different behavior than in previously published benchmarks, especially ICA-AROMA. These results demonstrate that a reproducible denoising benchmark can effectively assess the robustness of conclusions across multiple datasets and software versions. Technologies such as BIDS-App, the Jupyter Book and Neurolibre provided the infrastructure to publish the metadata and report figures. Readers can reproduce the report figures beyond the ones reported in the published manuscript. With the denoising benchmark, we hope to provide useful guidelines for the community, and that our software infrastructure will facilitate continued development as the state-of-the-art advances.

2022-12-31

(published)

www.semanticscholar.org

Re-Weighted Softmax Cross-Entropy to Control Forgetting in Federated Learning

Gwen Legate

Lucas Caccia

Eugene Belilovsky

In Federated Learning a global model is learned by aggregating model updates computed at a set of independent client nodes. To reduce commun… (see more)ication costs, multiple gradient steps are performed at each node prior to aggregation. A key challenge in this setting is data heterogeneity across clients resulting in differing local objectives. This can lead clients to overly minimize their own local objective consequently diverging from the global solution. We demonstrate that individual client models experience a catastrophic forgetting with respect to data from other clients and propose an efficient approach that modifies the cross-entropy objective on a per-client basis by re-weighting the softmax logits prior to computing the loss. This approach shields classes outside a client’s label set from abrupt representation change and we empirically demonstrate it can alleviate client forgetting and provide consistent improvements to standard federated learning algorithms. Our method is particularly beneficial under the most challenging federated learning settings where data heterogeneity is high and client participation in each round is low.

2022-12-31

Conference on Lifelong Learning Agents (published)

doi.org

proceedings.mlr.press

Robust and Versatile Bipedal Jumping Control through Multi-Task Reinforcement Learning

Zhongyu Li

Xue Bin Peng

Pieter Abbeel

Sergey Levine

Glen Berseth

Koushil Sreenath

2022-12-31

arXiv.org (preprint)

doi.org

Scalar Invariant Networks with Zero Bias

Chuqin Geng

Xiaojie Xu

Haolin Ye

Xujie Si

Just like weights, bias terms are the learnable parameters of many popular machine learning models, including neural networks. Biases are th… (see more)ought to enhance the representational power of neural networks, enabling them to solve a variety of tasks in computer vision. However, we argue that biases can be disregarded for some image-related tasks such as image classification, by considering the intrinsic distribution of images in the input space and desired model properties from first principles. Our findings suggest that zero-bias neural networks can perform comparably to biased networks for practical image classification tasks. We demonstrate that zero-bias neural networks possess a valuable property called scalar (multiplication) invariance. This means that the prediction of the network remains unchanged when the contrast of the input image is altered. We extend scalar invariance to more general cases, enabling formal verification of certain convex regions of the input space. Additionally, we prove that zero-bias neural networks are fair in predicting the zero image. Unlike state-of-the-art models that may exhibit bias toward certain labels, zero-bias networks have uniform belief in all labels. We believe dropping bias terms can be considered as a geometric prior in designing neural network architecture for image classification, which shares the spirit of adapting convolutions as the transnational invariance prior. The robustness and fairness advantages of zero-bias neural networks may also indicate a promising path towards trustworthy and ethical AI.

2022-12-31

NeurReps (published)

doi.org

proceedings.mlr.press

Scaling Self-Supervised End-to-End Driving with Multi-View Attention Learning

Yi Xiao

Felipe Codevilla

Diego Porres

Antonio M. López

2022-12-31

arXiv.org (preprint)

doi.org