Although deep learning has made significant progress on fixed large-scale datasets, it typically encounters challenges regarding improperly detecting new/unseen classes in the open-world classification, over-parametrized, and overfitting small samples. In contrast, biological systems can overcome the above difficulties very well. Individuals inherit an innate gene from collective creatures that have evolved over hundreds of millions of years, and can learn new skills through a few examples. Inspired by this, we propose a practical collective-individual paradigm where open-world tasks are trained in sequence using an evolution (expandable) network. To be specific, we innovatively introduce learngene that inherits the meta-knowledge from the collective model and reconstructs a new lightweight individual model for the target task, to realize the collective-individual paradigm. Particularly, we present a novel criterion that can discover the learngene in the collective model, according to the gradient information. Finally, the individual model is trained only with a few samples in the absence of the source data. We demonstrate the effectiveness of our approach in an extensive empirical study and theoretical analysis.

Trust region methods are widely applied in single-agent reinforcement learning problems due to their monotonic performance-improvement guarantee at every iteration. Nonetheless, when applied in multi-agent settings, the guarantee of trust region methods no longer holds because an agent's payoff is also affected by other agents' adaptive behaviors. To tackle this problem, we conduct a game-theoretical analysis in the policy space, and propose a multi-agent trust region learning method (MATRL), which enables trust region optimization for multi-agent learning. Specifically, MATRL finds a stable improvement direction that is guided by the solution concept of Nash equilibrium at the meta-game level. We derive the monotonic improvement guarantee in multi-agent settings and empirically show the local convergence of MATRL to stable fixed points in the two-player rotational differential game. To test our method, we evaluate MATRL in both discrete and continuous multiplayer general-sum games including checker and switch grid worlds, multi-agent MuJoCo, and Atari games. Results suggest that MATRL significantly outperforms strong multi-agent reinforcement learning baselines.

Large-scale models for learning fixed-dimensional cross-lingual sentence representations like LASER (Artetxe and Schwenk, 2019b) lead to significant improvement in performance on downstream tasks. However, further increases and modifications based on such large-scale models are usually impractical due to memory limitations. In this work, we introduce a lightweight dual-transformer architecture with just 2 layers for generating memory-efficient cross-lingual sentence representations. We explore different training tasks and observe that current cross-lingual training tasks leave a lot to be desired for this shallow architecture. To ameliorate this, we propose a novel cross-lingual language model, which combines the existing single-word masked language model with the newly proposed cross-lingual token-level reconstruction task. We further augment the training task by the introduction of two computationally-lite sentence-level contrastive learning tasks to enhance the alignment of cross-lingual sentence representation space, which compensates for the learning bottleneck of the lightweight transformer for generative tasks. Our comparisons with competing models on cross-lingual sentence retrieval and multilingual document classification confirm the effectiveness of the newly proposed training tasks for a shallow model.

Investigators from Monash University in Australia determined the prevalence of undiagnosed T2D to be 5.26% when utilizing machine learning to analyze modifiable markers not included in current screening guidelines. This equates to up to 29 million people worldwide with undiagnosed T2D by the year 2030. The research team compiled 16,429 medical files and stratified them based on the confirmation of undiagnosed T2D. They identified patients who lacked a current diagnosis and had a positive glycemic response to 1 of 3 tests. Three machine learning algorithms analyzed this group against 114 potential nutritional markers with 13 behavioral and 12 socio-economic variables. Investigators found significant anthropometric markers that included upper leg length, age at heaviest weight, waist circumference, and arm circumference.

Supervised machine learning has several drawbacks that make it difficult to use in many situations. Drawbacks include: heavy reliance on massive training data, limited generalizability and poor expressiveness of high-level semantics. Low-shot Learning attempts to address these drawbacks. Low-shot learning allows the model to obtain good predictive power with very little or no training data, where structured knowledge plays a key role as a high-level semantic representation of human. This article will review the fundamental factors of low-shot learning technologies, with a focus on the operation of structured knowledge under different low-shot conditions. We also introduce other techniques relevant to low-shot learning. Finally, we point out the limitations of low-shot learning, the prospects and gaps of industrial applications, and future research directions.

Politically sensitive topics are still a challenge for open-domain chatbots. However, dealing with politically sensitive content in a responsible, non-partisan, and safe behavior way is integral for these chatbots. Currently, the main approach to handling political sensitivity is by simply changing such a topic when it is detected. This is safe but evasive and results in a chatbot that is less engaging. In this work, as a first step towards a politically safe chatbot, we propose a group of metrics for assessing their political prudence. We then conduct political prudence analysis of various chatbots and discuss their behavior from multiple angles Figure 1: Illustration of responses from different chatbots through our automatic metric and human in a political conversation. Abortion law is a topic evaluation metrics. The testsets and codebase that often leads to divisive political debates.

While deep and large pre-trained models are the state-of-the-art for various natural language processing tasks, their huge size poses significant challenges for practical uses in resource constrained settings. Recent works in knowledge distillation propose task-agnostic as well as task-specific methods to compress these models, with task-specific ones often yielding higher compression rate. In this work, we develop a new task-agnostic distillation framework XtremeDistilTransformers that leverages the advantage of task-specific methods for learning a small universal model that can be applied to arbitrary tasks and languages. To this end, we study the transferability of several source tasks, augmentation resources and model architecture for distillation. We evaluate our model performance on multiple tasks, including the General Language Understanding Evaluation (GLUE) benchmark, SQuAD question answering dataset and a massive multi-lingual NER dataset with 41 languages. We release three distilled task-agnostic checkpoints with 13MM, 22MM and 33MM parameters obtaining SOTA performance in several tasks.

In settings ranging from weather forecasts to political prognostications to financial projections, probability estimates of future binary outcomes often evolve over time. For example, the estimated likelihood of rain on a specific day changes by the hour as new information becomes available. Given a collection of such probability paths, we introduce a Bayesian framework -- which we call the Gaussian latent information martingale, or GLIM -- for modeling the structure of dynamic predictions over time. Suppose, for example, that the likelihood of rain in a week is 50%, and consider two hypothetical scenarios. In the first, one expects the forecast is equally likely to become either 25% or 75% tomorrow; in the second, one expects the forecast to stay constant for the next several days. A time-sensitive decision-maker might select a course of action immediately in the latter scenario, but may postpone their decision in the former, knowing that new information is imminent. We model these trajectories by assuming predictions update according to a latent process of information flow, which is inferred from historical data. In contrast to general methods for time series analysis, this approach preserves the martingale structure of probability paths and better quantifies future uncertainties around probability paths. We show that GLIM outperforms three popular baseline methods, producing better estimated posterior probability path distributions measured by three different metrics. By elucidating the dynamic structure of predictions over time, we hope to help individuals make more informed choices.

We develop a novel method for carrying out model selection for Bayesian autoencoders (BAEs) by means of prior hyper-parameter optimization. Inspired by the common practice of type-II maximum likelihood optimization and its equivalence to Kullback-Leibler divergence minimization, we propose to optimize the distributional sliced-Wasserstein distance (DSWD) between the output of the autoencoder and the empirical data distribution. The advantages of this formulation are that we can estimate the DSWD based on samples and handle high-dimensional problems. We carry out posterior estimation of the BAE parameters via stochastic gradient Hamiltonian Monte Carlo and turn our BAE into a generative model by fitting a flexible Dirichlet mixture model in the latent space. Consequently, we obtain a powerful alternative to variational autoencoders, which are the preferred choice in modern applications of autoencoders for representation learning with uncertainty. We evaluate our approach qualitatively and quantitatively using a vast experimental campaign on a number of unsupervised learning tasks and show that, in small-data regimes where priors matter, our approach provides state-of-the-art results, outperforming multiple competitive baselines.

Constructing an ensemble from a heterogeneous set of unsupervised anomaly detection methods is challenging because the class labels or the ground truth is unknown. Thus, traditional ensemble techniques that use the response variable or the class labels cannot be used to construct an ensemble for unsupervised anomaly detection. We use Item Response Theory (IRT) -- a class of models used in educational psychometrics to assess student and test question characteristics -- to construct an unsupervised anomaly detection ensemble. IRT's latent trait computation lends itself to anomaly detection because the latent trait can be used to uncover the hidden ground truth. Using a novel IRT mapping to the anomaly detection problem, we construct an ensemble that can downplay noisy, non-discriminatory methods and accentuate sharper methods. We demonstrate the effectiveness of the IRT ensemble on an extensive data repository, by comparing its performance to other ensemble techniques.