More Like this

1. Meta-learning for Mixed Linear Regression

   Kong, Weihao ; Somani, Raghav ; Song, Zhao ; Kakade, Sham ; Oh, Sewoong ( July 2020 , International Conference on Machine Learning)

   null (Ed.)

   n modern supervised learning, there are a large number of tasks, but many of them are associated with only a small amount of labelled data. These include data from medical image processing and robotic interaction. Even though each individual task cannot be meaningfully trained in isolation, one seeks to meta-learn across the tasks from past experiences by exploiting some similarities. We study a fundamental question of interest: When can abundant tasks with small data compensate for lack of tasks with big data? We focus on a canonical scenario where each task is drawn from a mixture of k linear regressions, and identify sufficient conditions for such a graceful exchange to hold; there is little loss in sample complexity even when we only have access to small data tasks. To this end, we introduce a novel spectral approach and show that we can efficiently utilize small data tasks with the help of Omega(k^3/2) medium data tasks each with Omega(k^1/2) examples. 

   more » « less
2. Robust Meta-learning for Mixed Linear Regression with Small Batches

   Kong, Weihao ; Somani, Raghav ; Kakade, Sham ; Oh, Sewoong ( January 2020 , Advances in neural information processing systems)

   A common challenge faced in practical supervised learning, such as medical image processing and robotic interactions, is that there are plenty of tasks but each task cannot afford to collect enough labeled examples to be learned in isolation. However, by exploiting the similarities across those tasks, one can hope to overcome such data scarcity. Under a canonical scenario where each task is drawn from a mixture of k linear regressions, we study a fundamental question: can abundant small-data tasks compensate for the lack of big-data tasks? Existing second moment based approaches show that such a trade-off is efficiently achievable, with the help of medium-sized tasks with k^1/2 examples each. However, this algorithm is brittle in two important scenarios. The predictions can be arbitrarily bad even with only a few outliers in the dataset; or even if the medium-sized tasks are slightly smaller with. We introduce a spectral approach that is simultaneously robust under both scenarios. To this end, we first design a novel outlier-robust principal component analysis algorithm that achieves an optimal accuracy. This is followed by a sum-of-squares algorithm to exploit the information from higher order moments. Together, this approach is robust against outliers and achieves a graceful statistical trade-off. 

   more » « less
3. Matching Tasks and Workers under Known Arrival Distributions: Online Task Assignment with Two-sided Arrivals

   https://doi.org/10.1145/3652021  

   Dickerson, John P. ; Sankararaman, Karthik A. ; Srinivasan, Aravind ; Xu, Pan ; Xu, Yifan ( March 2024 , ACM Transactions on Economics and Computation)

   In this paper, we consider a setting inspired by spatial crowdsourcing platforms, where both workers and tasks arrive at different times, and each worker-task assignment yields a given reward. The key challenge is to address the uncertainty in the stochastic arrivals from both workers and the tasks. In this work, we consider a ubiquitous scenario where the arrival patterns of worker “types” and task “types” are not erratic but can be predicted from historical data. Specifically, we consider a finite time horizon T and assume that in each time-step the arrival of a worker and a task can be seen as an independent sample from two (different) distributions. Our model, called "Online Task Assignment with Two-Sided Arrival" (OTA-TSA), is a significant generalization of the classical online task-assignment problem when all the tasks are statically available. For the general case of OTA-TSA, we present an optimal non-adaptive algorithm (NADAP), which achieves a competitive ratio (CR) of at least 0.295. For a special case of OTA-TSA when the reward depends only on the worker type, we present two adaptive algorithms, which achieve CRs of at least 0.343 and 0.355, respectively. On the hardness side, we show that (1) no non-adaptive can achieve a CR larger than that of NADAP, establishing the optimality of NADAP among all non-adaptive algorithms; and (2) no (adaptive) algorithm can achieve a CR better than 0.581 (unconditionally) or 0.423 (conditionally on the benchmark linear program), respectively. All aforementioned negative results apply to even unweighted OTA-TSA when every assignment yields a uniform reward. At the heart of our analysis is a new technical tool, called "two-stage birth-death process", which is a refined notion of the classical birth-death process. We believe it may be of independent interest. Finally, we perform extensive numerical experiments on a real-world ride-share dataset collected in Chicago and a synthetic dataset, and results demonstrate the effectiveness of our proposed algorithms in practice. 

   more » « less
4. When BERT meets Bilbo: a learning curve analysis of pretrained language model on disease classification

   https://doi.org/10.1186/s12911-022-01829-2  

   Li, Xuedong ; Yuan, Walter ; Peng, Dezhong ; Mei, Qiaozhu ; Wang, Yue ( April 2022 , BMC Medical Informatics and Decision Making)

   Natural language processing (NLP) tasks in the health domain often deal with limited amount of labeled data due to high annotation costs and naturally rare observations. To compensate for the lack of training data, health NLP researchers often have to leverage knowledge and resources external to a task at hand. Recently, pretrained large-scale language models such as the Bidirectional Encoder Representations from Transformers (BERT) have been proven to be a powerful way of learning rich linguistic knowledge from massive unlabeled text and transferring that knowledge to downstream tasks. However, previous downstream tasks often used training data at such a large scale that is unlikely to obtain in the health domain. In this work, we aim to study whether BERT can still benefit downstream tasks when training data are relatively small in the context of health NLP.

   We conducted a learning curve analysis to study the behavior of BERT and baseline models as training data size increases. We observed the classification performance of these models on two disease diagnosis data sets, where some diseases are naturally rare and have very limited observations (fewer than 2 out of 10,000). The baselines included commonly used text classification models such as sparse and dense bag-of-words models, long short-term memory networks, and their variants that leveraged external knowledge. To obtain learning curves, we incremented the amount of training examples per disease from small to large, and measured the classification performance in macro-averaged$$F_{1}$$$F_1$score.

   On the task of classifying all diseases, the learning curves of BERT were consistently above all baselines, significantly outperforming them across the spectrum of training data sizes. But under extreme situations where only one or two training documents per disease were available, BERT was outperformed by linear classifiers with carefully engineered bag-of-words features.

   As long as the amount of training documents is not extremely few, fine-tuning a pretrained BERT model is a highly effective approach to health NLP tasks like disease classification. However, in extreme cases where each class has only one or two training documents and no more will be available, simple linear models using bag-of-words features shall be considered.

    

   more » « less
5. Enhanced Meta Reinforcement Learning using Demonstrations in Sparse Reward Environments

   Rengarajan, Desik ; Chaudhary, Sapana ; Jaewon, Kim ; Kalathil, Dileep ; Shakkottai, Srinivas ( November 2022 , Advances in neural information processing systems)

   Meta reinforcement learning (Meta-RL) is an approach wherein the experience gained from solving a variety of tasks is distilled into a meta-policy. The metapolicy, when adapted over only a small (or just a single) number of steps, is able to perform near-optimally on a new, related task. However, a major challenge to adopting this approach to solve real-world problems is that they are often associated with sparse reward functions that only indicate whether a task is completed partially or fully. We consider the situation where some data, possibly generated by a suboptimal agent, is available for each task. We then develop a class of algorithms entitled Enhanced Meta-RL using Demonstrations (EMRLD) that exploit this information—even if sub-optimal—to obtain guidance during training. We show how EMRLD jointly utilizes RL and supervised learning over the offline data to generate a meta-policy that demonstrates monotone performance improvements. We also develop a warm started variant called EMRLD-WS that is particularly efficient for sub-optimal demonstration data. Finally, we show that our EMRLD algorithms significantly outperform existing approaches in a variety of sparse reward environments, including that of a mobile robot. 

   more » « less