In high stake applications, active experimentation may be considered too risky and thus data are often collected passively. While in simple cases, such as in bandits, passive and active data collection are similarly effective, the price of passive sampling can be much higher when collecting data from a system with controlled states. The main focus of the current paper is the characterization of this price. For example, when learning in episodic finite state-action Markov decision processes (MDPs) with $\mathrm{S}$ states and $\mathrm{A}$ actions, we show that even with the best (but passively chosen) logging policy, $\Omega(\mathrm{A}^{\min(\mathrm{S}-1, H)}/\varepsilon^2)$ episodes are necessary (and sufficient) to obtain an $\epsilon$-optimal policy, where $H$ is the length of episodes. Note that this shows that the sample complexity blows up exponentially compared to the case of active data collection, a result which is not unexpected, but, as far as we know, have not been published beforehand and perhaps the form of the exact expression is a little surprising. We also extend these results in various directions, such as other criteria or learning in the presence of function approximation, with similar conclusions. A remarkable feature of our result is the sharp characterization of the exponent that appears, which is critical for understanding what makes passive learning hard.
相關內容
Secure computation protocols combine inputs from involved parties to generate an output while keeping their inputs private. Private Set Intersection (PSI) is a secure computation protocol that allows two parties, who each hold a set of items, to learn the intersection of their sets without revealing anything else about the items. Private Intersection Sum (PIS) extends PSI when the two parties want to learn the cardinality of the intersection, as well as the sum of the associated integer values for each identifier in the intersection, but nothing more. Finally, Private Join and Compute (PJC) is a scalable extension of PIS protocol to help organizations work together with confidential data sets. The extensions proposed in this paper include: (a) extending PJC protocol to additional data columns and applying columnar aggregation based on supported homomorphic operations, (b) exploring Ring Learning with Errors (RLWE) homomorphic encryption schemes to apply arithmetic operations such as sum and sum of squares, (c) ensuring stronger security using mutual authentication of communicating parties using certificates, and (d) developing a Website to operationalize such a service offering. We applied our results to develop a Proof-of-Concept solution called JingBing, a voter list validation service that allows different states to register, acquire secure communication modules, install it, and then conduct authenticated peer-to-peer communication. We conclude our paper with directions for future research to make such a solution scalable for practical real-life scenarios.
We introduce a new class of balanced allocation processes which are primarily characterized by ``filling'' underloaded bins. A prototypical example is the Packing process: At each round we only take one bin sample, if the load is below the average load, then we place as many balls until the average load is reached; otherwise, we place only one ball. We prove that for any process in this class the gap between the maximum and average load is $\mathcal{O}(\log n)$ w.h.p. for any number of balls $m\geq 1$. For the Packing process, we also provide a matching lower bound. Additionally, we prove that the Packing process is sample-efficient in the sense that the expected number of balls allocated per sample is strictly greater than one. Finally, we also demonstrate that the upper bound of $\mathcal{O}(\log n)$ on the gap can be extended to the Memory process studied by Mitzenmacher, Prabhakar and Shah (2002).
Mathematical methods are developed to characterize the asymptotics of recurrent neural networks (RNN) as the number of hidden units, data samples in the sequence, hidden state updates, and training steps simultaneously grow to infinity. In the case of an RNN with a simplified weight matrix, we prove the convergence of the RNN to the solution of an infinite-dimensional ODE coupled with the fixed point of a random algebraic equation. The analysis requires addressing several challenges which are unique to RNNs. In typical mean-field applications (e.g., feedforward neural networks), discrete updates are of magnitude $\mathcal{O}(\frac{1}{N})$ and the number of updates is $\mathcal{O}(N)$. Therefore, the system can be represented as an Euler approximation of an appropriate ODE/PDE, which it will converge to as $N \rightarrow \infty$. However, the RNN hidden layer updates are $\mathcal{O}(1)$. Therefore, RNNs cannot be represented as a discretization of an ODE/PDE and standard mean-field techniques cannot be applied. Instead, we develop a fixed point analysis for the evolution of the RNN memory states, with convergence estimates in terms of the number of update steps and the number of hidden units. The RNN hidden layer is studied as a function in a Sobolev space, whose evolution is governed by the data sequence (a Markov chain), the parameter updates, and its dependence on the RNN hidden layer at the previous time step. Due to the strong correlation between updates, a Poisson equation must be used to bound the fluctuations of the RNN around its limit equation. These mathematical methods give rise to the neural tangent kernel (NTK) limits for RNNs trained on data sequences as the number of data samples and size of the neural network grow to infinity.
With the rise of foundation models, a new artificial intelligence paradigm has emerged, by simply using general purpose foundation models with prompting to solve problems instead of training a separate machine learning model for each problem. Such models have been shown to have emergent properties of solving problems that they were not initially trained on. The studies for the effectiveness of such models are still quite limited. In this work, we widely study the capabilities of the ChatGPT models, namely GPT-4 and GPT-3.5, on 13 affective computing problems, namely aspect extraction, aspect polarity classification, opinion extraction, sentiment analysis, sentiment intensity ranking, emotions intensity ranking, suicide tendency detection, toxicity detection, well-being assessment, engagement measurement, personality assessment, sarcasm detection, and subjectivity detection. We introduce a framework to evaluate the ChatGPT models on regression-based problems, such as intensity ranking problems, by modelling them as pairwise ranking classification. We compare ChatGPT against more traditional NLP methods, such as end-to-end recurrent neural networks and transformers. The results demonstrate the emergent abilities of the ChatGPT models on a wide range of affective computing problems, where GPT-3.5 and especially GPT-4 have shown strong performance on many problems, particularly the ones related to sentiment, emotions, or toxicity. The ChatGPT models fell short for problems with implicit signals, such as engagement measurement and subjectivity detection.
The quality of text-to-image generation is continuously improving, yet the boundaries of its applicability are still unclear. In particular, refinement of the text input with the objective of achieving better results - commonly called prompt engineering - so far seems to have not been geared towards work with pre-existing texts. We investigate whether text-to-image generation and prompt engineering could be used to generate basic illustrations of popular fairytales. Using Midjourney v4, we engage in action research with a dual aim: to attempt to generate 5 believable illustrations for each of 5 popular fairytales, and to define a prompt engineering process that starts from a pre-existing text and arrives at an illustration of it. We arrive at a tentative 4-stage process: i) initial prompt, ii) composition adjustment, iii) style refinement, and iv) variation selection. We also discuss three reasons why the generation model struggles with certain illustrations: difficulties with counts, bias from stereotypical configurations and inability to depict overly fantastic situations. Our findings are not limited to the specific generation model and are intended to be generalisable to future ones.
With the advent of digital scanners and deep learning, diagnostic operations may move from a microscope to a desktop. Hematoxylin and Eosin (H&E) staining is one of the most frequently used stains for disease analysis, diagnosis, and grading, but pathologists do need different immunohistochemical (IHC) stains to analyze specific structures or cells. Obtaining all of these stains (H&E and different IHCs) on a single specimen is a tedious and time-consuming task. Consequently, virtual staining has emerged as an essential research direction. Here, we propose a novel generative model, Structural Cycle-GAN (SC-GAN), for synthesizing IHC stains from H&E images, and vice versa. Our method expressly incorporates structural information in the form of edges (in addition to color data) and employs attention modules exclusively in the decoder of the proposed generator model. This integration enhances feature localization and preserves contextual information during the generation process. In addition, a structural loss is incorporated to ensure accurate structure alignment between the generated and input markers. To demonstrate the efficacy of the proposed model, experiments are conducted with two IHC markers emphasizing distinct structures of glands in the colon: the nucleus of epithelial cells (CDX2) and the cytoplasm (CK818). Quantitative metrics such as FID and SSIM are frequently used for the analysis of generative models, but they do not correlate explicitly with higher-quality virtual staining results. Therefore, we propose two new quantitative metrics that correlate directly with the virtual staining specificity of IHC markers.
Image segmentation and depth estimation are crucial tasks in computer vision, especially in autonomous driving scenarios. Although these tasks are typically addressed separately, we propose an innovative approach to combine them in our novel deep learning network, Panoptic-DepthLab. By incorporating an additional depth estimation branch into the segmentation network, it can predict the depth of each instance segment. Evaluating on Cityscape dataset, we demonstrate the effectiveness of our method in achieving high-quality segmentation results with depth and visualize it with a color map. Our proposed method demonstrates a new possibility of combining different tasks and networks to generate a more comprehensive image recognition result to facilitate the safety of autonomous driving vehicles.
Shrinkage methods are frequently used to estimate fixed effects to reduce the noisiness of the least square estimators. However, widely used shrinkage estimators guarantee such noise reduction only under strong distributional assumptions. I develop an estimator for the fixed effects that obtains the best possible mean squared error within a class of shrinkage estimators. This class includes conventional shrinkage estimators and the optimality does not require distributional assumptions. The estimator has an intuitive form and is easy to implement. Moreover, the fixed effects are allowed to vary with time and to be serially correlated, and the shrinkage optimally incorporates the underlying correlation structure in this case. In such a context, I also provide a method to forecast fixed effects one period ahead.
Recent artificial intelligence (AI) systems have reached milestones in "grand challenges" ranging from Go to protein-folding. The capability to retrieve medical knowledge, reason over it, and answer medical questions comparably to physicians has long been viewed as one such grand challenge. Large language models (LLMs) have catalyzed significant progress in medical question answering; Med-PaLM was the first model to exceed a "passing" score in US Medical Licensing Examination (USMLE) style questions with a score of 67.2% on the MedQA dataset. However, this and other prior work suggested significant room for improvement, especially when models' answers were compared to clinicians' answers. Here we present Med-PaLM 2, which bridges these gaps by leveraging a combination of base LLM improvements (PaLM 2), medical domain finetuning, and prompting strategies including a novel ensemble refinement approach. Med-PaLM 2 scored up to 86.5% on the MedQA dataset, improving upon Med-PaLM by over 19% and setting a new state-of-the-art. We also observed performance approaching or exceeding state-of-the-art across MedMCQA, PubMedQA, and MMLU clinical topics datasets. We performed detailed human evaluations on long-form questions along multiple axes relevant to clinical applications. In pairwise comparative ranking of 1066 consumer medical questions, physicians preferred Med-PaLM 2 answers to those produced by physicians on eight of nine axes pertaining to clinical utility (p < 0.001). We also observed significant improvements compared to Med-PaLM on every evaluation axis (p < 0.001) on newly introduced datasets of 240 long-form "adversarial" questions to probe LLM limitations. While further studies are necessary to validate the efficacy of these models in real-world settings, these results highlight rapid progress towards physician-level performance in medical question answering.
Graph Neural Networks (GNNs) have shown promising results on a broad spectrum of applications. Most empirical studies of GNNs directly take the observed graph as input, assuming the observed structure perfectly depicts the accurate and complete relations between nodes. However, graphs in the real world are inevitably noisy or incomplete, which could even exacerbate the quality of graph representations. In this work, we propose a novel Variational Information Bottleneck guided Graph Structure Learning framework, namely VIB-GSL, in the perspective of information theory. VIB-GSL advances the Information Bottleneck (IB) principle for graph structure learning, providing a more elegant and universal framework for mining underlying task-relevant relations. VIB-GSL learns an informative and compressive graph structure to distill the actionable information for specific downstream tasks. VIB-GSL deduces a variational approximation for irregular graph data to form a tractable IB objective function, which facilitates training stability. Extensive experimental results demonstrate that the superior effectiveness and robustness of VIB-GSL.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191