Large language models have led to state-of-the-art accuracies across a range of tasks. However,training large language model needs massive computing resource, as more and more open source pre-training models are available, it is worthy to study how to take full advantage of available model. We find a method to save training time and resource cost by changing the small well-trained model to large model. We initialize a larger target model from a smaller source model by copy weight values from source model and padding with zeros or small initialization values on it to make the source and target model have approximate outputs, which is valid due to block matrix multiplication and residual connection in transformer structure. We test the target model on several data sets and find it is still comparable with the source model. When we continue training the target model, the training loss can start from a smaller value.
Due to the fact that basic uncertain information provides a simple form for decision information with certainty degree, it has been developed to reflect the quality of observed or subjective assessments. In order to study the algebra structure and preference relation of basic uncertain information, we develop some algebra operations for basic uncertain information. The order relation of such type of information has also been considered. Finally, to apply the developed algebra operations and order relations, a generalized TODIM method for multi-attribute decision making with basic uncertain information is given. The numerical example shows that the developed decision procedure is valid.
Ecosystem services are the direct and indirect contributions of an ecosystem to human well-being and survival. Ecosystem valuation is a method of assigning a monetary value to an ecosystem with its goods and services,often referred to as ecosystem service value (ESV). With the rapid expansion of cities, a mismatch occurs between urban development and ecological development, and it is increasingly urgent to establish a valid ecological assessment method. In this study, we propose an ecological evaluation standard framework by designing an ESV monitoring workflow based on the establishment of multi-level grids. The proposed method is able to capture multi-scale features, facilitates multi-level spatial expression, and can effectively reveal the spatial heterogeneity of ESV. Taking Haian city in the Jiangsu province as the study case, we implemented the proposed dynamic multi-level grids-based (DMLG) to calculate its urban ESV in 2016 and 2019. We found that the ESV of Haian city showed considerable growth (increased by 24.54 million RMB). Negative ESVs are concentrated in the central city, which presented a rapid trend of outward expansion. The results illustrated that the ongoing urban expanse does not reduce the ecological value in the study area. The proposed unified grid framework can be applied to other geographical regions and is expected to benefit future studies in ecosystem service evaluation in terms of capture multi-level spatial heterogeneity.
Large language models have shown promising results in zero-shot settings (Brown et al.,2020; Radford et al., 2019). For example, they can perform multiple choice tasks simply by conditioning on a question and selecting the answer with the highest probability. However, ranking by string probability can be problematic due to surface form competition-wherein different surface forms compete for probability mass, even if they represent the same underlying concept, e.g. "computer" and "PC." Since probability mass is finite, this lowers the probability of the correct answer, due to competition from other strings that are valid answers (but not one of the multiple choice options). We introduce Domain Conditional Pointwise Mutual Information, an alternative scoring function that directly compensates for surface form competition by simply reweighing each option according to a term that is proportional to its a priori likelihood within the context of the specific zero-shot task. It achieves consistent gains in zero-shot performance over both calibrated (Zhao et al., 2021) and uncalibrated scoring functions on all GPT-2 and GPT-3 models over a variety of multiple choice datasets.
Recently, deep generative models for molecular graphs are gaining more and more attention in the field of de novo drug design. A variety of models have been developed to generate topological structures of drug-like molecules, but explorations in generating three-dimensional structures are still limited. Existing methods have either focused on low molecular weight compounds without considering drug-likeness or generate 3D structures indirectly using atom density maps. In this work, we introduce Ligand Neural Network (L-Net), a novel graph generative model for designing drug-like molecules with high-quality 3D structures. L-Net directly outputs the topological and 3D structure of molecules (including hydrogen atoms), without the need for additional atom placement or bond order inference algorithm. The architecture of L-Net is specifically optimized for drug-like molecules, and a set of metrics is assembled to comprehensively evaluate its performance. The results show that L-Net is capable of generating chemically correct, conformationally valid, and highly druglike molecules. Finally, to demonstrate its potential in structure-based molecular design, we combine L-Net with MCTS and test its ability to generate potential inhibitors targeting ABL1 kinase.
We present a method for contraction-based feedback motion planning of locally incrementally exponentially stabilizable systems with unknown dynamics that provides probabilistic safety and reachability guarantees. Given a dynamics dataset, our method learns a deep control-affine approximation of the dynamics. To find a trusted domain where this model can be used for planning, we obtain an estimate of the Lipschitz constant of the model error, which is valid with a given probability, in a region around the training data, providing a local, spatially-varying model error bound. We derive a trajectory tracking error bound for a contraction-based controller that is subjected to this model error, and then learn a controller that optimizes this tracking bound. With a given probability, we verify the correctness of the controller and tracking error bound in the trusted domain. We then use the trajectory error bound together with the trusted domain to guide a sampling-based planner to return trajectories that can be robustly tracked in execution. We show results on a 4D car, a 6D quadrotor, and a 22D deformable object manipulation task, showing our method plans safely with learned models of high-dimensional underactuated systems, while baselines that plan without considering the tracking error bound or the trusted domain can fail to stabilize the system and become unsafe.
Sketch recognition algorithms are engineered and evaluated using publicly available datasets contributed by the sketch recognition community over the years. While existing datasets contain sketches of a limited set of generic objects, each new domain inevitably requires collecting new data for training domain specific recognizers. This gives rise to two fundamental concerns: First, will the data collection protocol yield ecologically valid data? Second, will the amount of collected data suffice to train sufficiently accurate classifiers? In this paper, we draw attention to these two concerns. We show that the ecological validity of the data collection protocol and the ability to accommodate small datasets are significant factors impacting recognizer accuracy in realistic scenarios. More specifically, using sketch-based gaming as a use case, we show that deep learning methods, as well as more traditional methods, suffer significantly from dataset shift. Furthermore, we demonstrate that in realistic scenarios where data is scarce and expensive, standard measures taken for adapting deep learners to small datasets fall short of comparing favorably with alternatives. Although transfer learning, and extensive data augmentation help deep learners, they still perform significantly worse compared to standard setups (e.g., SVMs and GBMs with standard feature representations). We pose learning from small datasets as a key problem for the deep sketch recognition field, one which has been ignored in the bulk of the existing literature.
We study multi-answer retrieval, an under-explored problem that requires retrieving passages to cover multiple distinct answers for a given question. This task requires joint modeling of retrieved passages, as models should not repeatedly retrieve passages containing the same answer at the cost of missing a different valid answer. In this paper, we introduce JPR, the first joint passage retrieval model for multi-answer retrieval. JPR makes use of an autoregressive reranker that selects a sequence of passages, each conditioned on previously selected passages. JPR is trained to select passages that cover new answers at each timestep and uses a tree-decoding algorithm to enable flexibility in the degree of diversity. Compared to prior approaches, JPR achieves significantly better answer coverage on three multi-answer datasets. When combined with downstream question answering, the improved retrieval enables larger answer generation models since they need to consider fewer passages, establishing a new state-of-the-art.
Our goal, in the context of open-domain textual question-answering (QA), is to explain answers by showing the line of reasoning from what is known to the answer, rather than simply showing a fragment of textual evidence (a "rationale'"). If this could be done, new opportunities for understanding and debugging the system's reasoning become possible. Our approach is to generate explanations in the form of entailment trees, namely a tree of multipremise entailment steps from facts that are known, through intermediate conclusions, to the hypothesis of interest (namely the question + answer). To train a model with this skill, we created ENTAILMENTBANK, the first dataset to contain multistep entailment trees. Given a hypothesis (question + answer), we define three increasingly difficult explanation tasks: generate a valid entailment tree given (a) all relevant sentences (b) all relevant and some irrelevant sentences, or (c) a corpus. We show that a strong language model can partially solve these tasks, in particular when the relevant sentences are included in the input (e.g., 35% of trees for (a) are perfect), and with indications of generalization to other domains. This work is significant as it provides a new type of dataset (multistep entailments) and baselines, offering a new avenue for the community to generate richer, more systematic explanations.
This paper studies the construction of p-values for nonparametric outlier detection, taking a multiple-testing perspective. The goal is to test whether new independent samples belong to the same distribution as a reference data set or are outliers. We propose a solution based on conformal inference, a broadly applicable framework which yields p-values that are marginally valid but mutually dependent for different test points. We prove these p-values are positively dependent and enable exact false discovery rate control, although in a relatively weak marginal sense. We then introduce a new method to compute p-values that are both valid conditionally on the training data and independent of each other for different test points; this paves the way to stronger type-I error guarantees. Our results depart from classical conformal inference as we leverage concentration inequalities rather than combinatorial arguments to establish our finite-sample guarantees. Furthermore, our techniques also yield a uniform confidence bound for the false positive rate of any outlier detection algorithm, as a function of the threshold applied to its raw statistics. Finally, the relevance of our results is demonstrated by numerical experiments on real and simulated data.
Scripts - standardized event sequences describing typical everyday activities - have been shown to help understand narratives by providing expectations, resolving ambiguity, and filling in unstated information. However, to date they have proved hard to author or extract from text. In this work, we demonstrate for the first time that pre-trained neural language models (LMs) can be be finetuned to generate high-quality scripts, at varying levels of granularity, for a wide range of everyday scenarios (e.g., bake a cake). To do this, we collected a large (6.4k), crowdsourced partially ordered scripts (named proScript), which is substantially larger than prior datasets, and developed models that generate scripts with combining language generation and structure prediction. We define two complementary tasks: (i) edge prediction: given a scenario and unordered events, organize the events into a valid (possibly partial-order) script, and (ii) script generation: given only a scenario, generate events and organize them into a (possibly partial-order) script. Our experiments show that our models perform well (e.g., F1=75.7 in task (i)), illustrating a new approach to overcoming previous barriers to script collection. We also show that there is still significant room for improvement toward human level performance. Together, our tasks, dataset, and models offer a new research direction for learning script knowledge.
Explainable AI (XAI) is a research area whose objective is to increase trustworthiness and to enlighten the hidden mechanism of opaque machine learning techniques. This becomes increasingly important in case such models are applied to the chemistry domain, for its potential impact on humans' health, e.g, toxicity analysis in pharmacology. In this paper, we present a novel approach to tackle explainability of deep graph networks in the context of molecule property prediction t asks, named MEG (Molecular Explanation Generator). We generate informative counterfactual explanations for a specific prediction under the form of (valid) compounds with high structural similarity and different predicted properties. Given a trained DGN, we train a reinforcement learning based generator to output counterfactual explanations. At each step, MEG feeds the current candidate counterfactual into the DGN, collects the prediction and uses it to reward the RL agent to guide the exploration. Furthermore, we restrict the action space of the agent in order to only keep actions that maintain the molecule in a valid state. We discuss the results showing how the model can convey non-ML experts with key insights into the learning model focus in the neighbourhood of a molecule.
Multiple instance learning (MIL) has become the standard learning paradigm for distantly supervised relation extraction (DSRE). However, due to relation extraction being performed at bag level, MIL has significant hardware requirements for training when coupled with large sentence encoders such as deep transformer neural networks. In this paper, we propose a novel sampling method for DSRE that relaxes these hardware requirements. In the proposed method, we limit the number of sentences in a batch by randomly sampling sentences from the bags in the batch. However, this comes at the cost of losing valid sentences from bags. To alleviate the issues caused by random sampling, we use an ensemble of trained models for prediction. We demonstrate the effectiveness of our approach by using our proposed learning setting to fine-tuning BERT on the widely NYT dataset. Our approach significantly outperforms previous state-of-the-art methods in terms of AUC and P@N metrics.
In this paper, we develop an analytical channel model for intelligent reflecting surface (IRS)-assisted free space optical (FSO) systems. Unlike IRS-assisted radio frequency systems, where it is typically assumed that a plane wave is incident on the IRS, in FSO systems, the incident wave is a Gaussian beam with non-uniform power distribution across the IRS. Taking this property into account, we develop an analytical end-to-end channel model for IRS-assisted FSO systems based on the Huygens-Fresnel principle. Our analytical model reveals the impact of the size, position, orientation, and phase-shift configuration of the IRS on the end-to-end channel. Furthermore, we show that results obtained based on geometric optics under the far-field approximation are only valid for a specific range of IRS-receiver lens distances depending on the IRS size, incident beam width, and wavelength. Simulation results validate the accuracy of the proposed analytical results for the FSO beam reflected from the IRS and compare the bit error rate performance obtained for the proposed analytical channel model with that obtained for geometric optics under the far-field approximation.
Applying compressive sensing (CS) allows for sub-Nyquist sampling in several application areas in 5G and beyond. This reduces the associated training, feedback, and computation overheads in many applications. However, the applicability of CS relies on the validity of a signal sparsity assumption and knowing the exact sparsity level. It is customary to assume a foreknown sparsity level. Still, this assumption is not valid in practice, especially when applying learned dictionaries as sparsifying transforms. The problem is more strongly pronounced with multidimensional sparsity. In this paper, we propose an algorithm for estimating the composite sparsity lying in multiple domains defined by learned dictionaries. The proposed algorithm estimates the sparsity level over a dictionary by inferring it from its counterpart with respect to a compact discrete Fourier basis. This inference is achieved by a machine learning prediction. This setting learns the intrinsic relationship between the columns of a dictionary and those of such a fixed basis. The proposed algorithm is applied to estimating sparsity levels in wireless channels, and in cognitive radio spectra. Extensive simulations validate a high quality of sparsity estimation leading to performances very close to the impractical case of assuming known sparsity.
Consider a robust multiple-input single-output downlink beamforming optimization problem in a frequency division duplexing system. The base station (BS) sends training signals to the users, and every user estimates the channel coefficients, quantizes the gain and the direction of the estimated channel and sends them back to the BS. Suppose that the channel state information at the transmitter is imperfectly known mainly due to the channel direction quantization errors, channel estimation errors and outdated channel effects. The actual channel is modeled as in an uncertainty set composed of two inequality homogeneous and one equality inhomogeneous quadratic constraints, in order to account for the aforementioned errors and effects. Then the transmit power minimization problem is formulated subject to robust signal-to-noise-plus-interference ratio constraints. Each robust constraint is transformed equivalently into a quadratic matrix inequality (QMI) constraint with respect to the beamforming vectors. The transformation is accomplished by an equivalent phase rotation process and the strong duality result for a quadratically constrained quadratic program. The minimization problem is accordingly turned into a QMI problem, and the problem is solved by a restricted linear matrix inequality relaxation with additional valid convex constraints. Simulation results are presented to demonstrate the performance of the proposed method, and show the efficiency of the restricted relaxation.
A new methodology for the design of isophoric thinned arrays with a priori controlled pattern features is introduced. A fully analytical and general (i.e., valid for any lattice and set of weights) relationship between the autocorrelation of the array excitations and the power pattern samples is first derived. Binary 2-D sequences with known autocorrelation properties, namely the difference sets (DSs), are then chosen as a representative benchmark to prove that it is possible to deduce closed-form synthesis formulas that a priori guarantee to fit requirements on the sidelobe level (SLL), the directivity, the half-power beamwidth, and the power pattern in user-defined directions. The selected results from a wide numerical assessment, which also includes full-wave simulations with realistic radiators, are illustrated to validate the reliability and the accuracy of the proposed design equations and the associated performance bounds.
In this paper, I reply to the recent article by Jeffery and Verheijen (2020) 'A new soil health policy paradigm: Pay for practice not performance!'. While expressing support for their call for a more pronounced role of soil protection in agri-environmental policy, I critically discuss the two main elements of their specific proposal: its emphasis of the concept of soil health and the recommendation to use action-based payments as the main policy instrument. I argue for using soil functions as a more established concept (and thus more adequate for policy purposes), which is also informationally richer than soil health. Furthermore, I provide a more differentiated discussion of the relative advantages and disadvantages of result-based and action-based payments, while addressing the specific criticisms towards the former that Jeffery and Verheijen voice. Also, I suggest an alternative approach (a hybrid model-based scheme) that addresses the limitations of both Jeffery and Verheijen's own proposal and the valid criticisms they direct at result-based payments.
Virtual character animation control is a problem for which Reinforcement Learning (RL) is a viable approach. While current work have applied RL effectively to portray physics-based skills, social behaviours are challenging to design reward functions for, due to their lack of physical interaction with the world. On the other hand, data-driven implementations for these skills have been limited to supervised learning methods which require extensive training data and carry constraints on generalisability. In this paper, we propose RLAnimate, a novel data-driven deep RL approach to address this challenge, where we combine the strengths of RL together with an ability to learn from a motion dataset when creating agents. We formalise a mathematical structure for training agents by refining the conceptual roles of elements such as agents, environments, states and actions, in a way that leverages attributes of the character animation domain and model-based RL. An agent trained using our approach learns versatile animation dynamics to portray multiple behaviours, using an iterative RL training process, which becomes aware of valid behaviours via representations learnt from motion capture clips. We demonstrate, by training agents that portray realistic pointing and waving behaviours, that our approach requires a significantly lower training time, and substantially fewer sample episodes to be generated during training relative to state-of-the-art physics-based RL methods. Also, compared to existing supervised learning-based animation agents, RLAnimate needs a limited dataset of motion clips to generate representations of valid behaviours during training.
Many decisions involve choosing an uncertain course of actions in deep and wide decision trees, as when we plan to visit an exotic country for vacation. In these cases, exhaustive search for the best sequence of actions is not tractable due to the large number of possibilities and limited time or computational resources available to make the decision. Therefore, planning agents need to balance breadth (exploring many actions at each level of the tree) and depth (exploring many levels in the tree) to allocate optimally their finite search capacity. We provide efficient analytical solutions and numerical analysis to the problem of allocating finite sampling capacity in one shot to large decision trees. We find that in general the optimal policy is to allocate few samples per level so that deep levels can be reached, thus favoring depth over breadth search. In contrast, in poor environments and at low capacity, it is best to broadly sample branches at the cost of not sampling deeply, although this policy is marginally better than deep allocations. Our results provide a theoretical foundation for the optimality of deep imagination for planning and show that it is a generally valid heuristic that could have evolved from the finite constraints of cognitive systems.