Search

Principal component analysis (PCA) is often used to analyze multivariate data together with cluster analysis, which depends on the number of principal components used. It is therefore important to determine the number of significant principal components (PCs) extracted from a data set. Here we use a variational Bayesian version of classical PCA, to develop a new method for estimating the number of significant PCs in contexts where the number of samples is of a similar to or greater than the number of features. This eliminates guesswork and potential bias in manually determining the number of principal components and avoids overestimation of variance by filtering noise. This framework can be applied to datasets of different shapes (number of rows and columns), different data types (binary, ordinal, categorical, continuous), and with noisy and missing data. Therefore, it is especially useful for data with arbitrary encodings and similar numbers of rows and columns, such as cultural, ecological, morphological, and behavioral datasets. We tested our method on both synthetic data and empirical datasets and found that it may underestimate but not overestimate the number of principal components for the synthetic data. A small number of components was found for each empirical dataset. These results suggest that it is broadly applicable across the life sciences., Comment: 9 pages, 2 figures

Deep neural networks drive the success of natural language processing. A fundamental property of language is its compositional structure, allowing humans to systematically produce forms for new meanings. For humans, languages with more compositional and transparent structures are typically easier to learn than those with opaque and irregular structures. However, this learnability advantage has not yet been shown for deep neural networks, limiting their use as models for human language learning. Here, we directly test how neural networks compare to humans in learning and generalizing different languages that vary in their degree of compositional structure. We evaluate the memorization and generalization capabilities of a large language model and recurrent neural networks, and show that both deep neural networks exhibit a learnability advantage for more structured linguistic input: neural networks exposed to more compositional languages show more systematic generalization, greater agreement between different agents, and greater similarity to human learners., Comment: 20 pages + supplementary material

Emergent communication protocols among humans and artificial neural network agents do not yet share the same properties and show some critical mismatches in results. We describe three important phenomena with respect to the emergence and benefits of compositionality: ease-of-learning, generalization, and group size effects (i.e., larger groups create more systematic languages). The latter two are not fully replicated with neural agents, which hinders the use of neural emergent communication for language evolution research. We argue that one possible reason for these mismatches is that key cognitive and communicative constraints of humans are not yet integrated. Specifically, in humans, memory constraints and the alternation between the roles of speaker and listener underlie the emergence of linguistic structure, yet these constraints are typically absent in neural simulations. We suggest that introducing such communicative and cognitive constraints would promote more linguistically plausible behaviors with neural agents., Comment: Published as a workshop paper at EmeCom at ICLR 2022

Computer science offers a large set of tools for prototyping, writing, running, testing, validating, sharing and reproducing results, however computational science lags behind. In the best case, authors may provide their source code as a compressed archive and they may feel confident their research is reproducible. But this is not exactly true. James Buckheit and David Donoho proposed more than two decades ago that an article about computational results is advertising, not scholarship. The actual scholarship is the full software environment, code, and data that produced the result. This implies new workflows, in particular in peer-reviews. Existing journals have been slow to adapt: source codes are rarely requested, hardly ever actually executed to check that they produce the results advertised in the article. ReScience is a peer-reviewed journal that targets computational research and encourages the explicit replication of already published research, promoting new and open-source implementations in order to ensure that the original research can be replicated from its description. To achieve this goal, the whole publishing chain is radically different from other traditional scientific journals. ReScience resides on GitHub where each new implementation of a computational study is made available together with comments, explanations, and software tests., Comment: 8 pages, 1 figure

Fisher developed his geometric model to support the micro-mutationalism hypothesis which claims that small mutations are more likely to be beneficial and therefore to contribute to evolution and adaptation. While others have provided a general solution to the model using geometric approaches, we derive an equivalent general solution using a probabilistic approach. Our approach to Fisher's geometric model provides alternative intuition and interpretation of the solution in terms of the model's parameters: for mutation to improve a phenotype, its relative beneficial effect must be larger than the ratio of its total effect and twice the difference between the current phenotype and the optimal one. Our approach provides new insight into this classical model of adaptive evolution., Comment: Post-print

We developed a smartphone application, Ha'Midgam, to poll and forecast the results of the 2015 Israeli elections. The application was downloaded by over 7,500 people. We present the method used to control bias in our sample and our forecasts. We discuss limitations of our approach and suggest possible solutions to control bias in similar applications., Comment: 5 pages, 1 table

Because mutations are mostly deleterious, mutation rates should be reduced by natural selection. However, mutations also provide the raw material for adaptation. Therefore, evolutionary theory suggests that the mutation rate must balance between adaptability - the ability to adapt - and adaptedness - the ability to remain adapted. We model an asexual population crossing a fitness valley and analyze the rate of complex adaptation with and without stress-induced mutagenesis - the increase of mutation rates in response to stress or maladaptation. We show that stress-induced mutagenesis increases the rate of complex adaptation without reducing the population mean fitness, thus breaking the evolutionary trade-off between adaptability and adaptedness. Our theoretical results support the hypothesis that stress-induced mutagenesis promotes adaptation and provide quantitative predictions of the rate of complex adaptation with different mutational strategies., Comment: 45 pages, 4 figures, 6 suppurting figures, 6 appendices