Adversarial vulnerabilities of human decision-making.

Adversarial examples are carefully crafted input patterns that are surprisingly poorly classified by artificial and/or natural neural networks. Here we examine adversarial vulnerabilities in the processes responsible for learning and choice in humans. Building upon recent recurrent neural network models of choice processes, we propose a general framework for generating adversarial opponents that can shape the choices of individuals in particular decision-making tasks toward the behavioral patterns desired by the adversary. We show the efficacy of the framework through three experiments involving action selection, response inhibition, and social decision-making. We further investigate the strategy used by the adversary in order to gain insights into the vulnerabilities of human choice. The framework may find applications across behavioral sciences in helping detect and avoid flawed choice.

A Geometric Clustering Tool (AGCT) to robustly unravel the inner cluster structures of time-series gene expressions.

Systems biology aims at holistically understanding the complexity of biological systems. In particular, nowadays with the broad availability of gene expression measurements, systems biology challenges the deciphering of the genetic cell machinery from them. In order to help researchers, reverse engineer the genetic cell machinery from these noisy datasets, interactive exploratory clustering methods, pipelines and gene clustering tools have to be specifically developed. Prior methods/tools for time series data, however, do not have the following four major ingredients in analytic and methodological view point: (i) principled time-series feature extraction methods, (ii) variety of manifold learning methods for capturing high-level view of the dataset, (iii) high-end automatic structure extraction, and (iv) friendliness to the biological user community. With a view to meet the requirements, we present AGCT (A Geometric Clustering Tool), a software package used to unravel the complex architecture of large-scale, non-necessarily synchronized time-series gene expression data. AGCT capture signals on exhaustive wavelet expansions of the data, which are then embedded on a low-dimensional non-linear map using manifold learning algorithms, where geometric proximity captures potential interactions. Post-processing techniques, including hard and soft information geometric clustering algorithms, facilitate the summarizing of the complete map as a smaller number of principal factors which can then be formally identified using embedded statistical inference techniques. Three-dimension interactive visualization and scenario recording over the processing helps to reproduce data analysis results without additional time. Analysis of the whole-cell Yeast Metabolic Cycle (YMC) moreover, Yeast Cell Cycle (YCC) datasets demonstrate AGCT's ability to accurately dissect all stages of metabolism and the cell cycle progression, independently of the time course and the number of patterns related to the signal. Analysis of Pentachlorophenol iduced dataset demonstrat how AGCT dissects data to identify two networks: Interferon signaling and NRF2-signaling networks.

Optimal response vigor and choice under non-stationary outcome values.

Within a rational framework, a decision-maker selects actions based on the reward-maximization principle, which stipulates that they acquire outcomes with the highest value at the lowest cost. Action selection can be divided into two dimensions: selecting an action from various alternatives, and choosing its vigor, i.e., how fast the selected action should be executed. Both of these dimensions depend on the values of outcomes, which are often affected as more outcomes are consumed together with their associated actions. Despite this, previous research has only addressed the computational substrate of optimal actions in the specific condition that the values of outcomes are constant. It is not known what actions are optimal when the values of outcomes are non-stationary. Here, based on an optimal control framework, we derive a computational model for optimal actions when outcome values are non-stationary. The results imply that, even when the values of outcomes are changing, the optimal response rate is constant rather than decreasing. This finding shows that, in contrast to previous theories, commonly observed changes in action rate cannot be attributed solely to changes in outcome value. We then prove that this observation can be explained based on uncertainty about temporal horizons; e.g., the session duration. We further show that, when multiple outcomes are available, the model explains probability matching as well as maximization strategies. The model therefore provides a quantitative analysis of optimal action and explicit predictions for future testing.

An Advanced Omic Approach to Identify Co-Regulated Clusters and Transcription Regulation Network with AGCT and SHOE Methods.

To obtain the global picture of genetic machinery for massive high-throughput gene expression data, novel data-driven unsupervised learning approaches are becoming essentially important. For this purpose, basic analytic workflow has been established and should include two steps: first, unsupervised clustering to identify genes with similar behavior upon exposure to a signal, and second, identification of transcription factors regulating those genes. In this chapter, we will describe an advanced tool that can be used for analyzing and characterizing large-scale time-series gene expression composed of a two-step approach. For the first step, we developed an original method "A Geometric Clustering Tool" (AGCT) that unveils the complex architecture of large-scale time-series gene expression data in a real-time manner using cutting edge techniques of low dimension manifold learning, data clustering, and visualization. For the second step, we established an original method "Sequence Homology in Eukaryotes" (SHOE) executing comparative genomic analysis on humans, mice, and rats.

Gentle Nearest Neighbors Boosting over Proper Scoring Rules.

Tailoring nearest neighbors algorithms to boosting is an important problem. Recent papers study an approach, UNN, which provably minimizes particular convex surrogates under weak assumptions. However, numerical issues make it necessary to experimentally tweak parts of the UNN algorithm, at the possible expense of the algorithm's convergence and performance. In this paper, we propose a lightweight Newton-Raphson alternative optimizing proper scoring rules from a very broad set, and establish formal convergence rates under the boosting framework that compete with those known for UNN. To the best of our knowledge, no such boosting-compliant convergence rates were previously known in the popular Gentle Adaboost's lineage. We provide experiments on a dozen domains, including Caltech and SUN computer vision databases, comparing our approach to major families including support vector machines, (Ada)boosting and stochastic gradient descent. They support three major conclusions: (i) GNNB significantly outperforms UNN, in terms of convergence rate and quality of the outputs, (ii) GNNB performs on par with or better than computationally intensive large margin approaches, (iii) on large domains that rule out those latter approaches for computational reasons, GNNB provides a simple and competitive contender to stochastic gradient descent. Experiments include a divide-and-conquer improvement of GNNB exploiting the link with proper scoring rules optimization.

High-resolution random-modulation cw lidar.

A high-resolution random-modulation continuous wave lidar for surface detection using a semiconductor laser diode is presented. The laser diode is intensity modulated with the pseudorandom binary sequence. Its enhanced resolution is achieved via interpolation and a novel front-end analog technique, lowering the requirement of the analog-to-digital converter sampling rate and the associated circuitry. Its mathematical model is presented, including the derivation of the signal-to-noise ratio and the distance standard deviation. Analytical and experimental results demonstrate its capability to achieve distance accuracy of less than 2 cm within 2.6 ms acquisition time, over distances ranging from 1 to 12 m. The laser diode emits 1.4 mW of optical power at a wavelength of 635 nm.

Bregman divergences and surrogates for learning.

Bartlett et al. (2006) recently proved that a ground condition for surrogates, classification calibration, ties up their consistent minimization to that of the classification risk, and left as an important problem the algorithmic questions about their minimization. In this paper, we address this problem for a wide set which lies at the intersection of classification calibrated surrogates and those of Murata et al. (2004). This set coincides with those satisfying three common assumptions about surrogates. Equivalent expressions for the members-sometimes well known-follow for convex and concave surrogates, frequently used in the induction of linear separators and decision trees. Most notably, they share remarkable algorithmic features: for each of these two types of classifiers, we give a minimization algorithm provably converging to the minimum of any such surrogate. While seemingly different, we show that these algorithms are offshoots of the same "master" algorithm. This provides a new and broad unified account of different popular algorithms, including additive regression with the squared loss, the logistic loss, and the top-down induction performed in CART, C4.5. Moreover, we show that the induction enjoys the most popular boosting features, regardless of the surrogate. Experiments are provided on 40 readily available domains.

G-protein coupled receptor signaling architecture of mammalian immune cells.

A series of recent studies on large-scale networks of signaling and metabolic systems revealed that a certain network structure often called "bow-tie network" are observed. In signaling systems, bow-tie network takes a form with diverse and redundant inputs and outputs connected via a small numbers of core molecules. While arguments have been made that such network architecture enhances robustness and evolvability of biological systems, its functional role at a cellular level remains obscure. A hypothesis was proposed that such a network function as a stimuli-reaction classifier where dynamics of core molecules dictate downstream transcriptional activities, hence physiological responses against stimuli. In this study, we examined whether such hypothesis can be verified using experimental data from Alliance for Cellular Signaling (AfCS) that comprehensively measured GPCR related ligands response for B-cell and macrophage. In a GPCR signaling system, cAMP and Ca2+ act as core molecules. Stimuli-response for 32 ligands to B-Cells and 23 ligands to macrophages has been measured. We found that ligands with correlated changes of cAMP and Ca2+ tend to cluster closely together within the hyperspaces of both cell types and they induced genes involved in the same cellular processes. It was found that ligands inducing cAMP synthesis activate genes involved in cell growth and proliferation; cAMP and Ca2+ molecules that increased together form a feedback loop and induce immune cells to migrate and adhere together. In contrast, ligands without a core molecules response are scattered throughout the hyperspace and do not share clusters. G-protein coupling receptors together with immune response specific receptors were found in cAMP and Ca2+ activated clusters. Analyses have been done on the original software applicable for discovering 'bow-tie' network architectures within the complex network of intracellular signaling where ab initio clustering has been implemented as well. Groups of potential transcription factors for each specific group of genes were found to be partly conserved across B-cell and macrophage. A series of findings support the hypothesis.

On weighting clustering.

Recent papers and patents in iterative unsupervised learning have emphasized a new trend in clustering. It basically consists of penalizing solutions via weights on the instance points, somehow making clustering move toward the hardest points to cluster. The motivations come principally from an analogy with powerful supervised classification methods known as boosting algorithms. However, interest in this analogy has so far been mainly borne out from experimental studies only. This paper is, to the best of our knowledge, the first attempt at its formalization. More precisely, we handle clustering as a constrained minimization of a Bregman divergence. Weight modifications rely on the local variations of the expected complete log-likelihoods. Theoretical results show benefits resembling those of boosting algorithms and bring modified (weighted) versions of clustering algorithms such as k-means, fuzzy c-means, Expectation Maximization (EM), and k-harmonic means. Experiments are provided for all these algorithms, with a readily available code. They display the advantages that subtle data reweighting may bring to clustering.

Statistical region merging.

This paper explores a statistical basis for a process often described in computer vision: image segmentation by region merging following a particular order in the choice of regions. We exhibit a particular blend of algorithmics and statistics whose segmentation error is, as we show, limited from both the qualitative and quantitative standpoints. This approach can be efficiently approximated in linear time/space, leading to a fast segmentation algorithm tailored to processing images described using most common numerical pixel attribute spaces. The conceptual simplicity of the approach makes it simple to modify and cope with hard noise corruption, handle occlusion, authorize the control of the segmentation scale, and process unconventional data such as spherical images. Experiments on gray-level and color images, obtained with a short readily available C-code, display the quality of the segmentations obtained.