The quantification of cognitive powers rests on identifying a behavioural task that depends on them. Such dependence cannot be assured, for the powers a task invokes cannot be experimentally controlled or constrained a priori, resulting in unknown vulnerability to failure of specificity and generalisability. Evaluating a compact version of Raven's Advanced Progressive Matrices (RAPM), a widely used clinical test of fluid intelligence, we show that LaMa, a self-supervised artificial neural network trained solely on the completion of partially masked images of natural environmental scenes, achieves human-level test scores a prima vista, without any task-specific inductive bias or training. Compared with cohorts of healthy and focally lesioned participants, LaMa exhibits human-like variation with item difficulty, and produces errors characteristic of right frontal lobe damage under degradation of its ability to integrate global spatial patterns. LaMa's narrow training and limited capacity -- comparable to the nervous system of the fruit fly -- suggest RAPM may be open to computationally simple solutions that need not necessarily invoke abstract reasoning.
相關內容
The equioscillation condition is extended to multivariate approximation. To this end, it is reformulated as the synchronized oscillations between the error maximizers and the components of a related Haar matrix kernel vector. This new condition gives rise to a multivariate equioscillation theorem where the Haar condition is not assumed and hence the existence and the characterization by equioscillation become independent of uniqueness. This allows the theorem to be applicable to problems with no strong uniqueness or even no uniqueness. A technical additional requirement on the involved Haar matrix and its kernel vector is proved to be sufficient for strong uniqueness. Instances of multivariate problems with strongly unique, unique and nonunique solutions are presented to illustrate the scope of the theorem.
Expecting intelligent machines to efficiently work in real world requires a new method to understand unstructured information in unknown environments with good accuracy, scalability and generalization, like human. Here, a memristive neural computing based perceptual signal differential processing and learning method for intelligent machines is presented, via extracting main features of environmental information and applying associated encoded stimuli to memristors, we successfully obtain human-like ability in processing unstructured environmental information, such as amplification (>720%) and adaptation (<50%) of mechanical stimuli. The method also exhibits good scalability and generalization, validated in two typical applications of intelligent machines: object grasping and autonomous driving. In the former, a robot hand experimentally realizes safe and stable grasping, through learning unknown object features (e.g., sharp corner and smooth surface) with a single memristor in 1 ms. In the latter, the decision-making information of 10 unstructured environments in autonomous driving (e.g., overtaking cars, pedestrians) are accurately (94%) extracted with a 40x25 memristor array. By mimicking the intrinsic nature of human low-level perception mechanisms in electronic memristive neural circuits, the proposed method is adaptable to diverse sensing technologies, helping intelligent machines to generate smart high-level decisions in real world.
We show a relation, based on parallel repetition of the Magic Square game, that can be solved, with probability exponentially close to $1$ (worst-case input), by $1D$ (uniform) depth $2$, geometrically-local, noisy (noise below a threshold), fan-in $4$, quantum circuits. We show that the same relation cannot be solved, with an exponentially small success probability (averaged over inputs drawn uniformly), by $1D$ (non-uniform) geometrically-local, sub-linear depth, classical circuits consisting of fan-in $2$ NAND gates. Quantum and classical circuits are allowed to use input-independent (geometrically-non-local) resource states, that is entanglement and randomness respectively. To the best of our knowledge, previous best (analogous) depth separation for a task between quantum and classical circuits was constant v/s sub-logarithmic, although for general (geometrically non-local) circuits. Our hardness result for classical circuits is based on a direct product theorem about classical communication protocols from Jain and Kundu [JK22]. As an application, we propose a protocol that can potentially demonstrate verifiable quantum advantage in the NISQ era. We also provide generalizations of our result for higher dimensional circuits as well as a wider class of Bell games.
An overview is presented of a general theory of statistical inference that is referred to as the fiducial-Bayes fusion. This theory combines organic fiducial inference and Bayesian inference. The aim is that the reader is given a clear summary of the conceptual framework of the fiducial-Bayes fusion as well as pointers to further reading about its more technical aspects. Particular attention is paid to the issue of how much importance should be attached to the role of Bayesian inference within this framework. The appendix contains a substantive example of the application of the theory of the fiducial-Bayes fusion, which supplements various other examples of the application of this theory that are referenced in the paper.
Generalized cross-validation (GCV) is a widely-used method for estimating the squared out-of-sample prediction risk that employs a scalar degrees of freedom adjustment (in a multiplicative sense) to the squared training error. In this paper, we examine the consistency of GCV for estimating the prediction risk of arbitrary ensembles of penalized least squares estimators. We show that GCV is inconsistent for any finite ensemble of size greater than one. Towards repairing this shortcoming, we identify a correction that involves an additional scalar correction (in an additive sense) based on degrees of freedom adjusted training errors from each ensemble component. The proposed estimator (termed CGCV) maintains the computational advantages of GCV and requires neither sample splitting, model refitting, or out-of-bag risk estimation. The estimator stems from a finer inspection of ensemble risk decomposition and two intermediate risk estimators for the components in this decomposition. We provide a non-asymptotic analysis of the CGCV and the two intermediate risk estimators for ensembles of convex penalized estimators under Gaussian features and a linear response model. In the special case of ridge regression, we extend the analysis to general feature and response distributions using random matrix theory, which establishes model-free uniform consistency of CGCV.
What can be learned about causality and experimentation from passive data? This question is salient given recent successes of passively-trained language models in interactive domains such as tool use. Passive learning is inherently limited. However, we show that purely passive learning can in fact allow an agent to learn generalizable strategies for determining and using causal structures, as long as the agent can intervene at test time. We formally illustrate that learning a strategy of first experimenting, then seeking goals, can allow generalization from passive learning in principle. We then show empirically that agents trained via imitation on expert data can indeed generalize at test time to infer and use causal links which are never present in the training data; these agents can also generalize experimentation strategies to novel variable sets never observed in training. We then show that strategies for causal intervention and exploitation can be generalized from passive data even in a more complex environment with high-dimensional observations, with the support of natural language explanations. Explanations can even allow passive learners to generalize out-of-distribution from perfectly-confounded training data. Finally, we show that language models, trained only on passive next-word prediction, can generalize causal intervention strategies from a few-shot prompt containing examples of experimentation, together with explanations and reasoning. These results highlight the surprising power of passive learning of active causal strategies, and may help to understand the behaviors and capabilities of language models.
Meta-analysis aims to combine effect measures from several studies. For continuous outcomes, the most popular effect measures use simple or standardized differences in sample means. However, a number of applications focus on the absolute values of these effect measures (i.e., unsigned magnitude effects). We provide statistical methods for meta-analysis of magnitude effects based on standardized mean differences. We propose a suitable statistical model for random-effects meta-analysis of absolute standardized mean differences (ASMD), investigate a number of statistical methods for point and interval estimation, and provide practical recommendations for choosing among them.
Complexity is a fundamental concept underlying statistical learning theory that aims to inform generalization performance. Parameter count, while successful in low-dimensional settings, is not well-justified for overparameterized settings when the number of parameters is more than the number of training samples. We revisit complexity measures based on Rissanen's principle of minimum description length (MDL) and define a novel MDL-based complexity (MDL-COMP) that remains valid for overparameterized models. MDL-COMP is defined via an optimality criterion over the encodings induced by a good Ridge estimator class. We provide an extensive theoretical characterization of MDL-COMP for linear models and kernel methods and show that it is not just a function of parameter count, but rather a function of the singular values of the design or the kernel matrix and the signal-to-noise ratio. For a linear model with $n$ observations, $d$ parameters, and i.i.d. Gaussian predictors, MDL-COMP scales linearly with $d$ when $d<n$, but the scaling is exponentially smaller -- $\log d$ for $d>n$. For kernel methods, we show that MDL-COMP informs minimax in-sample error, and can decrease as the dimensionality of the input increases. We also prove that MDL-COMP upper bounds the in-sample mean squared error (MSE). Via an array of simulations and real-data experiments, we show that a data-driven Prac-MDL-COMP informs hyper-parameter tuning for optimizing test MSE with ridge regression in limited data settings, sometimes improving upon cross-validation and (always) saving computational costs. Finally, our findings also suggest that the recently observed double decent phenomenons in overparameterized models might be a consequence of the choice of non-ideal estimators.
Problems from metric graph theory such as Metric Dimension, Geodetic Set, and Strong Metric Dimension have recently had a strong impact on the field of parameterized complexity by being the first problems in NP to admit double-exponential lower bounds in the treewidth, and even in the vertex cover number for the latter. We initiate the study of enumerating minimal solution sets for these problems and show that they are also of great interest in enumeration. More specifically, we show that enumerating minimal resolving sets in graphs and minimal geodetic sets in split graphs are equivalent to hypergraph dualization, arguably one of the most important open problems in algorithmic enumeration. This provides two new natural examples to a question that emerged in different works this last decade: for which vertex (or edge) set graph property $\Pi$ is the enumeration of minimal (or maximal) subsets satisfying $\Pi$ equivalent to hypergraph dualization? As only very few properties are known to fit within this context -- namely, properties related to minimal domination -- our results make significant progress in characterizing such properties, and provide new angles of approach for tackling hypergraph dualization. In a second step, we consider cases where our reductions do not apply, namely graphs with no long induced paths, and show these cases to be mainly tractable.
Forecasting has always been at the forefront of decision making and planning. The uncertainty that surrounds the future is both exciting and challenging, with individuals and organisations seeking to minimise risks and maximise utilities. The large number of forecasting applications calls for a diverse set of forecasting methods to tackle real-life challenges. This article provides a non-systematic review of the theory and the practice of forecasting. We provide an overview of a wide range of theoretical, state-of-the-art models, methods, principles, and approaches to prepare, produce, organise, and evaluate forecasts. We then demonstrate how such theoretical concepts are applied in a variety of real-life contexts. We do not claim that this review is an exhaustive list of methods and applications. However, we wish that our encyclopedic presentation will offer a point of reference for the rich work that has been undertaken over the last decades, with some key insights for the future of forecasting theory and practice. Given its encyclopedic nature, the intended mode of reading is non-linear. We offer cross-references to allow the readers to navigate through the various topics. We complement the theoretical concepts and applications covered by large lists of free or open-source software implementations and publicly-available databases.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191