Search

The challenges faced by neural networks on tabular data are well-documented and have hampered the progress of tabular foundation models. Techniques leveraging in-context learning (ICL) have shown promise here, allowing for dynamic adaptation to unseen data. ICL can provide predictions for entirely new datasets without further training or hyperparameter tuning, therefore providing very fast inference when encountering a novel task. However, scaling ICL for tabular data remains an issue: approaches based on large language models cannot efficiently process numeric tables, and tabular-specific techniques have not been able to effectively harness the power of real data to improve performance and generalization. We are able to overcome these challenges by training tabular-specific ICL-based architectures on real data with self-supervised learning and retrieval, combining the best of both worlds. Our resulting model -- the Tabular Discriminative Pre-trained Transformer (TabDPT) -- achieves state-of-the-art performance on the CC18 (classification) and CTR23 (regression) benchmarks with no task-specific fine-tuning, demonstrating the adapatability and speed of ICL once the model is pre-trained. TabDPT also demonstrates strong scaling as both model size and amount of available data increase, pointing towards future improvements simply through the curation of larger tabular pre-training datasets and training larger models., Comment: Minimal TabDPT interface to provide predictions on new datasets available at the following link: https://github.com/layer6ai-labs/TabDPT

Tabular data is a pervasive modality spanning a wide range of domains, and the inherent diversity poses a considerable challenge for deep learning. Recent advancements using transformer-based in-context learning have shown promise on smaller and less complex datasets, but have struggled to scale to larger and more complex ones. To address this limitation, we propose a combination of retrieval and fine-tuning: we can adapt the transformer to a local subset of the data by collecting nearest neighbours, and then perform task-specific fine-tuning with this retrieved set of neighbours in context. Using TabPFN as the base model -- currently the best tabular in-context learner -- and applying our retrieval and fine-tuning scheme on top results in what we call a locally-calibrated PFN, or LoCalPFN. We conduct extensive evaluation on 95 datasets curated by TabZilla from OpenML, upon which we establish a new state-of-the-art with LoCalPFN -- even with respect to tuned tree-based models. Notably, we show a significant boost in performance compared to the base in-context model, demonstrating the efficacy of our approach and advancing the frontier of deep learning in tabular data.

The goal of multimodal alignment is to learn a single latent space that is shared between multimodal inputs. The most powerful models in this space have been trained using massive datasets of paired inputs and large-scale computational resources, making them prohibitively expensive to train in many practical scenarios. We surmise that existing unimodal encoders pre-trained on large amounts of unimodal data should provide an effective bootstrap to create multimodal models from unimodal ones at much lower costs. We therefore propose FuseMix, a multimodal augmentation scheme that operates on the latent spaces of arbitrary pre-trained unimodal encoders. Using FuseMix for multimodal alignment, we achieve competitive performance -- and in certain cases outperform state-of-the art methods -- in both image-text and audio-text retrieval, with orders of magnitude less compute and data: for example, we outperform CLIP on the Flickr30K text-to-image retrieval task with $\sim \! 600\times$ fewer GPU days and $\sim \! 80\times$ fewer image-text pairs. Additionally, we show how our method can be applied to convert pre-trained text-to-image generative models into audio-to-image ones. Code is available at: https://github.com/layer6ai-labs/fusemix., Comment: CVPR 2024 (Highlight)

Transformer-based models have greatly pushed the boundaries of time series forecasting recently. Existing methods typically encode time series data into $\textit{patches}$ using one or a fixed set of patch lengths. This, however, could result in a lack of ability to capture the variety of intricate temporal dependencies present in real-world multi-periodic time series. In this paper, we propose MultiResFormer, which dynamically models temporal variations by adaptively choosing optimal patch lengths. Concretely, at the beginning of each layer, time series data is encoded into several parallel branches, each using a detected periodicity, before going through the transformer encoder block. We conduct extensive evaluations on long- and short-term forecasting datasets comparing MultiResFormer with state-of-the-art baselines. MultiResFormer outperforms patch-based Transformer baselines on long-term forecasting tasks and also consistently outperforms CNN baselines by a large margin, while using much fewer parameters than these baselines.

Self-supervised representation learning~(SSRL) has advanced considerably by exploiting the transformation invariance assumption under artificially designed data augmentations. While augmentation-based SSRL algorithms push the boundaries of performance in computer vision and natural language processing, they are often not directly applicable to other data modalities, and can conflict with application-specific data augmentation constraints. This paper presents an SSRL approach that can be applied to any data modality and network architecture because it does not rely on augmentations or masking. Specifically, we show that high-quality data representations can be learned by reconstructing random data projections. We evaluate the proposed approach on a wide range of representation learning tasks that span diverse modalities and real-world applications. We show that it outperforms multiple state-of-the-art SSRL baselines. Due to its wide applicability and strong empirical results, we argue that learning from randomness is a fruitful research direction worthy of attention and further study., Comment: Published as a conference paper of ICLR 2024. https://openreview.net/pdf?id=EpYnZpDpsQ

Electronic health records (EHRs) recorded in hospital settings typically contain a wide range of numeric time series data that is characterized by high sparsity and irregular observations. Effective modelling for such data must exploit its time series nature, the semantic relationship between different types of observations, and information in the sparsity structure of the data. Self-supervised Transformers have shown outstanding performance in a variety of structured tasks in NLP and computer vision. But multivariate time series data contains structured relationships over two dimensions: time and recorded event type, and straightforward applications of Transformers to time series data do not leverage this distinct structure. The quadratic scaling of self-attention layers can also significantly limit the input sequence length without appropriate input engineering. We introduce the DuETT architecture, an extension of Transformers designed to attend over both time and event type dimensions, yielding robust representations from EHR data. DuETT uses an aggregated input where sparse time series are transformed into a regular sequence with fixed length; this lowers the computational complexity relative to previous EHR Transformer models and, more importantly, enables the use of larger and deeper neural networks. When trained with self-supervised prediction tasks, that provide rich and informative signals for model pre-training, our model outperforms state-of-the-art deep learning models on multiple downstream tasks from the MIMIC-IV and PhysioNet-2012 EHR datasets., Comment: Accepted at MLHC 2023, camera-ready version

The computational benefits of iterative non-autoregressive transformers decrease as the number of decoding steps increases. As a remedy, we introduce Distill Multiple Steps (DiMS), a simple yet effective distillation technique to decrease the number of required steps to reach a certain translation quality. The distilled model enjoys the computational benefits of early iterations while preserving the enhancements from several iterative steps. DiMS relies on two models namely student and teacher. The student is optimized to predict the output of the teacher after multiple decoding steps while the teacher follows the student via a slow-moving average. The moving average keeps the teacher's knowledge updated and enhances the quality of the labels provided by the teacher. During inference, the student is used for translation and no additional computation is added. We verify the effectiveness of DiMS on various models obtaining 7.8 and 12.9 BLEU points improvements in single-step translation accuracy on distilled and raw versions of WMT'14 De-En., Comment: Code available at https://github.com/layer6ai-labs/DiMS

In text-video retrieval, the objective is to learn a cross-modal similarity function between a text and a video that ranks relevant text-video pairs higher than irrelevant pairs. However, videos inherently express a much wider gamut of information than texts. Instead, texts often capture sub-regions of entire videos and are most semantically similar to certain frames within videos. Therefore, for a given text, a retrieval model should focus on the text's most semantically similar video sub-regions to make a more relevant comparison. Yet, most existing works aggregate entire videos without directly considering text. Common text-agnostic aggregations schemes include mean-pooling or self-attention over the frames, but these are likely to encode misleading visual information not described in the given text. To address this, we propose a cross-modal attention model called X-Pool that reasons between a text and the frames of a video. Our core mechanism is a scaled dot product attention for a text to attend to its most semantically similar frames. We then generate an aggregated video representation conditioned on the text's attention weights over the frames. We evaluate our method on three benchmark datasets of MSR-VTT, MSVD and LSMDC, achieving new state-of-the-art results by up to 12% in relative improvement in Recall@1. Our findings thereby highlight the importance of joint text-video reasoning to extract important visual cues according to text. Full code and demo can be found at: https://layer6ai-labs.github.io/xpool/, Comment: CVPR 2022

Institutions in highly regulated domains such as finance and healthcare often have restrictive rules around data sharing. Federated learning is a distributed learning framework that enables multi-institutional collaborations on decentralized data with improved protection for each collaborator's data privacy. In this paper, we propose a communication-efficient scheme for decentralized federated learning called ProxyFL, or proxy-based federated learning. Each participant in ProxyFL maintains two models, a private model, and a publicly shared proxy model designed to protect the participant's privacy. Proxy models allow efficient information exchange among participants without the need of a centralized server. The proposed method eliminates a significant limitation of canonical federated learning by allowing model heterogeneity; each participant can have a private model with any architecture. Furthermore, our protocol for communication by proxy leads to stronger privacy guarantees using differential privacy analysis. Experiments on popular image datasets, and a cancer diagnostic problem using high-quality gigapixel histology whole slide images, show that ProxyFL can outperform existing alternatives with much less communication overhead and stronger privacy.

Time series data introduces two key challenges for explainability methods: firstly, observations of the same feature over subsequent time steps are not independent, and secondly, the same feature can have varying importance to model predictions over time. In this paper, we propose Windowed Feature Importance in Time (WinIT), a feature removal based explainability approach to address these issues. Unlike existing feature removal explanation methods, WinIT explicitly accounts for the temporal dependence between different observations of the same feature in the construction of its importance score. Furthermore, WinIT captures the varying importance of a feature over time, by summarizing its importance over a window of past time steps. We conduct an extensive empirical study on synthetic and real-world data, compare against a wide range of leading explainability methods, and explore the impact of various evaluation strategies. Our results show that WinIT achieves significant gains over existing methods, with more consistent performance across different evaluation metrics. The code for our work is publicly available at \url{https://github.com/layer6ai-labs/WinIT}., Comment: International Conference on Learning Representations 2023 (ICLR'23)

Localizing actions in video is a core task in computer vision. The weakly supervised temporal localization problem investigates whether this task can be adequately solved with only video-level labels, significantly reducing the amount of expensive and error-prone annotation that is required. A common approach is to train a frame-level classifier where frames with the highest class probability are selected to make a video-level prediction. Frame level activations are then used for localization. However, the absence of frame-level annotations cause the classifier to impart class bias on every frame. To address this, we propose the Action Selection Learning (ASL) approach to capture the general concept of action, a property we refer to as "actionness". Under ASL, the model is trained with a novel class-agnostic task to predict which frames will be selected by the classifier. Empirically, we show that ASL outperforms leading baselines on two popular benchmarks THUMOS-14 and ActivityNet-1.2, with 10.3% and 5.7% relative improvement respectively. We further analyze the properties of ASL and demonstrate the importance of actionness. Full code for this work is available here: https://github.com/layer6ai-labs/ASL., Comment: CVPR 2021

The fundamental question of how to best simulate quantum systems using conventional computational resources lies at the forefront of condensed matter and quantum computation. It impacts both our understanding of quantum materials and our ability to emulate quantum circuits. Here we present an exact formulation of quantum dynamics via factorized generalized measurements which maps quantum states to probability distributions with the advantage that local unitary dynamics and quantum channels map to local quasi-stochastic matrices. This representation provides a general framework for using state-of-the-art probabilistic models in machine learning for the simulation of quantum many-body dynamics. Using this framework, we have developed a practical algorithm to simulate quantum circuits with the Transformer, a powerful ansatz responsible for the most recent breakthroughs in natural language processing. We demonstrate our approach by simulating circuits which build GHZ and linear graph states of up to 60 qubits, as well as a variational quantum eigensolver circuit for preparing the ground state of the transverse field Ising model on six qubits. Our methodology constitutes a modern machine learning approach to the simulation of quantum physics with applicability both to quantum circuits as well as other quantum many-body systems., Comment: 12 pages, 11 figures including supplementary materials

We present our winning solution to the Open Images 2019 Visual Relationship challenge. This is the largest challenge of its kind to date with nearly 9 million training images. Challenge task consists of detecting objects and identifying relationships between them in complex scenes. Our solution has three stages, first object detection model is fine-tuned for the challenge classes using a novel weight transfer approach. Then, spatio-semantic and visual relationship models are trained on candidate object pairs. Finally, features and model predictions are combined to generate the final relationship prediction. Throughout the challenge we focused on minimizing the hardware requirements of our architecture. Specifically, our weight transfer approach enables much faster optimization, allowing the entire architecture to be trained on a single GPU in under two days. In addition to efficient optimization, our approach also achieves superior accuracy winning first place out of over 200 teams, and outperforming the second place team by over $5\%$ on the held-out private leaderboard.

We present a novel Cross-Class Relevance Learning approach for the task of temporal concept localization. Most localization architectures rely on feature extraction layers followed by a classification layer which outputs class probabilities for each segment. However, in many real-world applications classes can exhibit complex relationships that are difficult to model with this architecture. In contrast, we propose to incorporate target class and class-related features as input, and learn a pairwise binary model to predict general segment to class relevance. This facilitates learning of shared information between classes, and allows for arbitrary class-specific feature engineering. We apply this approach to the 3rd YouTube-8M Video Understanding Challenge together with other leading models, and achieve first place out of over 280 teams. In this paper we describe our approach and show some empirical results.

We present our solution to Landmark Image Retrieval Challenge 2019. This challenge was based on the large Google Landmarks Dataset V2[9]. The goal was to retrieve all database images containing the same landmark for every provided query image. Our solution is a combination of global and local models to form an initial KNN graph. We then use a novel extension of the recently proposed graph traversal method EGT [1] referred to as semi-supervised EGT to refine the graph and retrieve better candidates.

Leveraging health administrative data (HAD) datasets for predicting the risk of chronic diseases including diabetes has gained a lot of attention in the machine learning community recently. In this paper, we use the largest health records datasets of patients in Ontario,Canada. Provided by the Institute of Clinical Evaluative Sciences (ICES), this database is age, gender and ethnicity-diverse. The datasets include demographics, lab measurements,drug benefits, healthcare system interactions, ambulatory and hospitalizations records. We perform one of the first large-scale machine learning studies with this data to study the task of predicting diabetes in a range of 1-10 years ahead, which requires no additional screening of individuals.In the best setup, we reach a test AUC of 80.3 with a single-model trained on an observation window of 5 years with a one-year buffer using all datasets. A subset of top 15 features alone (out of a total of 963) could provide a test AUC of 79.1. In this paper, we provide extensive machine learning model performance and feature contribution analysis, which enables us to narrow down to the most important features useful for diabetes forecasting. Examples include chronic conditions such as asthma and hypertension, lab results, diagnostic codes in insurance claims, age and geographical information., Comment: 18 pages, 3 figures, 8 Tables, Submitted to 2019 ML for Healthcare conference

Previous highly scalable one-class collaborative filtering methods such as Projected Linear Recommendation (PLRec) have advocated using fast randomized SVD to embed items into a latent space, followed by linear regression methods to learn personalized recommendation models per user. Unfortunately, naive SVD embedding methods often exhibit a popularity bias that skews the ability to accurately embed niche items. To address this, we leverage insights from Noise Contrastive Estimation (NCE) to derive a closed-form, efficiently computable "depopularized" embedding. While this method is not ideal for direct recommendation using methods like PureSVD since popularity still plays an important role in recommendation, we find that embedding followed by linear regression to learn personalized user models in a novel method we call NCE-PLRec leverages the improved item embedding of NCE while correcting for its popularity unbiasing in final recommendations. An analysis of the recommendation popularity distribution demonstrates that NCE-PLRec uniformly distributes its recommendations over the popularity spectrum while other methods exhibit distinct biases towards specific popularity subranges, thus artificially restricting their recommendations. Empirically, NCE-PLRec outperforms state-of-the-art methods as well as various ablations of itself on a variety of large-scale recommendation datasets., Comment: 8 pages

We propose a new model for unsupervised document embedding. Leading existing approaches either require complex inference or use recurrent neural networks (RNN) that are difficult to parallelize. We take a different route and develop a convolutional neural network (CNN) embedding model. Our CNN architecture is fully parallelizable resulting in over 10x speedup in inference time over RNN models. Parallelizable architecture enables to train deeper models where each successive layer has increasingly larger receptive field and models longer range semantic structure within the document. We additionally propose a fully unsupervised learning algorithm to train this model based on stochastic forward prediction. Empirical results on two public benchmarks show that our approach produces comparable to state-of-the-art accuracy at a fraction of computational cost., Comment: Major revision with additional experiments and model description

We present our solution to the Yandex Personalized Web Search Challenge. The aim of this challenge was to use the historical search logs to personalize top-N document rankings for a set of test users. We used over 100 features extracted from user- and query-depended contexts to train neural net and tree-based learning-to-rank and regression models. Our final submission, which was a blend of several different models, achieved an NDCG@10 of 0.80476 and placed 4'th amongst the 194 teams winning 3'rd prize.

We consider the problem of training probabilistic conditional random fields (CRFs) in the context of a task where performance is measured using a specific loss function. While maximum likelihood is the most common approach to training CRFs, it ignores the inherent structure of the task's loss function. We describe alternatives to maximum likelihood which take that loss into account. These include a novel adaptation of a loss upper bound from the structured SVMs literature to the CRF context, as well as a new loss-inspired KL divergence objective which relies on the probabilistic nature of CRFs. These loss-sensitive objectives are compared to maximum likelihood using ranking as a benchmark task. This comparison confirms the importance of incorporating loss information in the probabilistic training of CRFs, with the loss-inspired KL outperforming all other objectives.