United States

Automatic International Classification of Diseases (ICD) coding aims to assign multiple ICD codes to a medical note with an average of 3,000+ tokens. This task is challenging due to the high-dimensional space of multi-label assignment (155,000+ ICD code candidates) and the long-tail challenge - Many ICD codes are infrequently assigned yet infrequent ICD codes are important clinically. This study addresses the long-tail challenge by transforming this multi-label classification task into an autoregressive generation task. Specifically, we first introduce a novel pretraining objective to generate free text diagnosis and procedure descriptions using the SOAP structure, the medical logic physicians use for note documentation. Second, instead of directly predicting the high dimensional space of ICD codes, our model generates the lower dimension of text descriptions, which then infer ICD codes. Third, we designed a novel prompt template for multi-label classification. We evaluate our Generation with Prompt (GP) model with the benchmark of all code assignment (MIMIC-III-full) and few shot ICD code assignment evaluation benchmark (MIMIC-III-few). Experiments on MIMIC-III-few show that our model performs with a marco F1 30.2, which substantially outperforms the previous MIMIC-III-full SOTA model (marco F1 4.3) and the model specifically designed for few/zero shot setting (marco F1 18.7). Finally, we design a novel ensemble learner, a cross attention reranker with prompts, to integrate previous SOTA and our best few-shot coding predictions. Experiments on MIMIC-III-full show that our ensemble learner substantially improves both macro and micro F1, from 10.4 to 14.6 and from 58.2 to 59.1, respectively. 

AAAI 2023

Multi-label Few-shot ICD Coding as Autoregressive Generation with Prompt

snlp: text classification

ml: multi-class/multi-label learning & extreme classification

snlp: language models

snlp: applications

snlp: generation

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-23 is the Thirty-Seventh AAAI Conference on Artificial Intelligence. The theme of this conference is to create collaborative bridges within and beyond AI. Like previous AAAI conferences, AAAI-23 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and two new activities: a Bridge Program and a Lab Program. Many of these activities are tailored to the theme of bridges and all are selected according to the highest standards, with additional programs for students and young researchers. 
AAAI is providing you with a conference planner, which you can use to help organize your itinerary of activities. This includes talks to attend in person, talks to attend remotely, breaks with colleagues and your site seeing activities. To access this conference planner, please go to [https://aaai-2023.takemobi.io](https://aaai-2023.takemobi.io).

In order to access this site, you need to register. If you haven't already, please register [here](https://aaai.org/Conferences/AAAI-23/registration/).


The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines.

poster

This paper studies how to train machine-learning models that directly approximate the optimal solutions of constrained optimization problems. This is an empirical risk minimization under constraints, which is challenging as training must balance optimality and feasibility conditions. Supervised learning methods often approach this challenge by training the model on a large collection of pre-solved instances. This paper takes a different route and proposes the idea of Primal-Dual Learning (PDL), a self-supervised training method that does not require a set of pre-solved instances or an optimization solver for training and inference. Instead, PDL mimics the trajectory of an Augmented Lagrangian Method (ALM) and jointly trains primal and dual neural networks. Being a primal-dual method, PDL uses instance-specific penalties of the constraint terms in the loss function used to train the primal network. Experiments show that, on a set of nonlinear optimization benchmarks, PDL typically exhibits negligible constraint violations and minor optimality gaps, and is remarkably close to the ALM optimization. PDL also demonstrated improved or similar performance in terms of the optimality gaps, constraint violations, and training times compared to existing approaches.

Self-Supervised Primal-Dual Learning for Constrained Optimization

Training deep neural networks (DNNs) with noisy labels often leads to poorly generalized models as DNNs tend to memorize the noisy labels in training. Various strategies have been developed for improving sample selection precision and mitigating the noisy label memorization issue. However, most existing works adopt a class-dependent softmax classifier that is vulnerable to noisy labels by entangling the classification of multi-class features. This paper presents a class-independent regularization (CIR) method that can effectively alleviate the negative impact of noisy labels in DNN training. CIR regularizes the class-dependent softmax classifier by introducing multi-binary classifiers each of which takes care of one class only. Thanks to its class-independent nature, CIR is tolerant to noisy labels as misclassification by one binary classifier does not affect others. For effective training of CIR, we design a heterogeneous adaptive co-teaching strategy that forces the class-independent and class-dependent classifiers to focus on sample selection and image classification, respectively, in a cooperative manner. Extensive experiments show that CIR achieves superior performance consistently across multiple benchmarks with both synthetic and real images. Code is
available at https://github.com/RumengYi/CIR.

Class-Independent Regularization for Learning with Noisy Labels

Hyperspectral image (HSI) denoising is a crucial preprocessing procedure for the subsequent HSI applications. Unfortunately, though witnessing the development of deep learning in HSI denoising area, existing convolution-based methods face the trade-off between computational efficiency and capability to model non-local characteristics of HSI. In this paper, we propose a Spatial-Spectral Transformer (SST) to alleviate this problem. To fully explore intrinsic similarity characteristics in both spatial dimension and spectral dimension, we conduct non-local spatial self-attention and global spectral self-attention with Transformer architecture. The window-based spatial self-attention focuses on the spatial similarity beyond the neighboring region. While, the spectral self-attention exploits the long-range dependencies between highly correlative bands. Experimental results show that our proposed method outperforms the state-of-the-art HSI denoising methods in quantitative quality and visual results. The code is released at https://github.com/MyuLi/SST.

Spatial-Spectral Transformer for Hyperspectral Image Denoising

Are Federated Learning (FL) systems free from backdoor poisoning with the arsenal of various defense strategies deployed? This is an intriguing problem with significant practical implications regarding the utility of FL services. Despite the recent flourish of poisoning-resilient FL methods, our study shows that carefully tuning the collusion between malicious participants can minimize the trigger-induced bias of the poisoned local model from the poison-free one, which plays the key role in delivering stealthy backdoor attacks and circumventing a wide spectrum of state-of-the-art defense methods in FL. In our work, we instantiate the attack strategy by proposing a distributed backdoor attack method, namely Cerberus Poisoning (CerP). It jointly tunes the backdoor trigger and controls the poisoned model changes on each malicious participant to achieve a stealthy yet successful backdoor attack against a wide spectrum of defensive mechanisms of federated learning techniques. Our extensive study on 3 large-scale benchmark datasets and 13 mainstream defensive mechanisms confirms that Cerberus Poisoning raises a significantly severe threat to the integrity and security of federated learning practices, regardless of the flourish of robust Federated Learning methods.

Poisoning with Cerberus: Stealthy and Colluded Backdoor Attack against Federated Learning

Although multi-view clustering (MVC) has achieved remarkable performance by integrating the complementary information of views, it is inefficient when facing scalable data. Proverbially, anchor strategy can mitigate such a challenge a certain extent. However, the unsupervised dynamic strategy usually cannot obtain the optimal anchors for MVC. The main reasons are that it does not consider the fairness of different views and lacks the priori supervised guidance. To completely solve these problems, we first propose the priori anchor graph regularization (PAGG) for scalable multi-view bipartite graph clustering, dubbed as SMGC method. Specifically, SMGC learns a few representative consensus anchors to simulate the numerous view data well, and constructs a bipartite graph to bridge the affinities between the anchors and original data points. In order to largely improve the quality of anchors, PAGG predefines prior anchor labels to constrain the anchors with discriminative cluster structure and fair view allocation, such that a better bipartite graph can be obtained for fast clustering. Experimentally, abundant of experiments are accomplished on six scalable benchmark datasets, and the experimental results fully demonstrate the effectiveness and efficiency of our SMGC.

Priori Anchor Labels Supervised Scalable Multi-view Bipartite Graph Clustering

We consider the fair division problem of indivisible items. It is well-known that an envy-free allocation may not exist, and a relaxed version of envy-freeness, envy-freeness up to one item (EF1), has been widely considered. In an EF1 allocation, an agent may envy others' allocated shares, but only up to one item. In many applications, we may wish to specify a subset of prioritized agents where strict envy-freeness needs to be guaranteed from these agents to the remaining agents, while ensuring the whole allocation is still EF1. Prioritized agents may be those agents who are envious in a previous EF1 allocation, those agents who belong to underrepresented groups, etc. Motivated by this, we propose a new fairness notion named envy-freeness with prioritized agents EFprior, and study the existence and the algorithmic aspects for the problem of computing an EFprior allocation. With additive valuations, the simple round-robin algorithm is able to compute an EFprior allocation. In this paper, we mainly focus on general valuations. In particular, we present a polynomial-time algorithm that outputs an EFprior allocation with most of the items allocated. When all the items need to be allocated, we also present polynomial-time algorithms for some well-motivated special cases.

Fair Division with Prioritized Agents

Existing episodic reinforcement algorithms assume that the  length of an episode is fixed across time and known a priori. In this paper, we consider a general framework of  episodic reinforcement learning when the length of each episode is drawn from a distribution. We first establish that this problem is equivalent to online reinforcement learning with general discounting where the learner is trying to optimize the expected discounted sum of rewards over an infinite horizon, but where the discounting function is not necessarily geometric. We show that minimizing regret with this new general discounting is equivalent to minimizing regret with uncertain episode lengths. We then design a reinforcement learning algorithm that minimizes regret with general discounting but acts for the setting with uncertain episode lengths. We instantiate  our general bound for different types of discounting, including geometric and polynomial discounting. We also show that we can obtain similar regret bounds even when the uncertainty over the episode lengths is unknown, by estimating the unknown distribution over time. Finally, we compare our learning algorithms with existing value-iteration based episodic RL algorithms on a grid-world environment.

Online Reinforcement Learning with Uncertain Episode Lengths

In offline multi-agent reinforcement learning (MARL), the agents estimate policies from a given dataset. We study reward-poisoning attack on offline MARL, where an exogenous attacker can modify the reward vectors in the dataset before the agents see the data. Similar to poisoning attacks on a single RL agent, the attacker wants to guide the agents to learn a nefarious target policy while minimizing the lp norm of reward modification. Unlike attacks on single-agent RL, the attacker takes advantage of the strategic interactions between the agents to install the target policy as a Markov Perfect Dominant Strategy Equilibrium (MPDSE) and can achieve lower reward modification costs compared to separate single-agent attacks. We then show that the attack works on various MARL agents including uncertainty-aware learners, and we exhibit linear programs to efficiently solve the offline MARL attack problem. Our work thus paves the way for studying defense in offline MARL in the future.

Reward Poisoning Attacks on Offline Multi-Agent Reinforcement Learning

Given a possibly false claim sentence, how can we automatically correct it with minimal editing? Existing methods either require a large number of pairs of false and corrected claims for supervised training or do not handle well errors spanning over multiple tokens within an utterance. In this paper, we propose VENCE, a novel method for factual error correction (FEC) with minimal edits. VENCE formulates the FEC problem as iterative sampling editing actions with respect to a target density function. We carefully design the target function with predicted truthfulness scores from an offline trained fact verification model. VENCE samples the most probable editing positions based on back-calculated gradients of the truthfulness score concerning input tokens and the editing actions using a distantly-supervised language model (T5). Experiments on a public dataset show that VENCE improves the well-adopted SARI metric by 5.3 (or a relative improvement of 11.8%) over the previous best distantly-supervised methods.

Converge to the Truth: Factual Error Correction via Iterative Constrained Editing

Question answering (QA) models for reading comprehension tend to learn shortcut solutions rather than the solutions intended by QA datasets.
QA models that have learned shortcut solutions can achieve human-level performance in shortcut examples where shortcuts are valid, but these same behaviors degrade generalization potential on anti-shortcut examples where shortcuts are invalid.
Various methods have been proposed to mitigate this problem, but they do not fully take the characteristics of shortcuts themselves into account.
We assume that the learnability of shortcuts, i.e., how easy it is to learn a shortcut, is useful to mitigate the problem.
Thus, we first examine the learnability of the representative shortcuts on extractive and multiple-choice QA datasets.
Behavioral tests using biased training sets reveal that shortcuts that exploit answer positions and word-label correlations are preferentially learned for extractive and multiple-choice QA, respectively.
We find that the more learnable a shortcut is, the flatter and deeper the loss landscape is around the shortcut solution in the parameter space.
We also find that the availability of the preferred shortcuts tends to make the task easier to perform from an information-theoretic viewpoint.
Lastly, we experimentally show that the learnability of shortcuts can be utilized to construct an effective QA training set; the more learnable a shortcut is, the smaller the proportion of anti-shortcut examples required to achieve comparable performance on shortcut and anti-shortcut examples.
We claim that the learnability of shortcuts should be considered when designing mitigation methods.

Downloads

Next from AAAI 2023

Self-Supervised Primal-Dual Learning for Constrained Optimization

Similar lecture

Decoupling the Skeleton Parsing and Schema Linking for Text-to-SQL