In lifelong learning, data are used to improve performance not only on the present task, but also on past and future (unencountered) tasks. While typical transfer learning algorithms can improve performance on future tasks, their performance on prior tasks degrades upon learning new tasks (called forgetting). Many recent approaches for continual or lifelong learning have attempted to maintain performance on old tasks given new tasks. But striving to avoid forgetting sets the goal unnecessarily low. The goal of lifelong learning should be to use data to improve performance on both future tasks (forward transfer) and past tasks (backward transfer). In this paper, we show that a simple approach-representation ensembling-demonstrates both forward and backward transfer in a variety of simulated and benchmark data scenarios, including tabular, vision (CIFAR-100, 5-dataset, Split Mini-Imagenet, Food1k, and CORe50), and speech (spoken digit), in contrast to various reference algorithms, which typically failed to transfer either forward or backward, or both. Moreover, our proposed approach can flexibly operate with or without a computational budget.

Tree ensembles, random forests and gradient boosted trees, are useful in resource-limited machine learning deployments. However, traversing tree data structures is not cache friendly, which results in high latency during inference or regression. Tree traversal incurs random I/Os making inference memory bound. We present a system that trades many random I/Os for few sequential I/O by remapping a forest of trees into a single spatial index. It builds on the observation that each leaf in the forest encodes a hyperrectangle in the feature space. We make queries I/O efficient through pruning and space-filling curves. We then optimize computation through quantization of hyperrectangle boundaries and vectorization of enclosure queries. Our evaluation on a diverse set of benchmark datasets shows that the system reduces inference latency by 2 times in memory and 10 times for external memory with no detectable loss of accuracy.

In real-world applications, the distribution of the data, and our goals, evolve over time. The prevailing theoretical framework for studying machine learning, namely probably approximately correct (PAC) learning, largely ignores time. As a consequence, existing strategies to address the dynamic nature of data and goals exhibit poor real-world performance. This paper develops a theoretical framework called "Prospective Learning" that is tailored for situations when the optimal hypothesis changes over time. In PAC learning, empirical risk minimization (ERM) is known to be consistent. We develop a learner called Prospective ERM, which returns a sequence of predictors that make predictions on future data. We prove that the risk of prospective ERM converges to the Bayes risk under certain assumptions on the stochastic process generating the data. Prospective ERM, roughly speaking, incorporates time as an input in addition to the data. We show that standard ERM as done in PAC learning, without incorporating time, can result in failure to learn when distributions are dynamic. Numerical experiments illustrate that prospective ERM can learn synthetic and visual recognition problems constructed from MNIST and CIFAR-10. Code at https://github.com/neurodata/prolearn.

Why do brains have inhibitory connections? Why do deep networks have negative weights? We propose an answer from the perspective of representation capacity. We believe representing functions is the primary role of both (i) the brain in natural intelligence, and (ii) deep networks in artificial intelligence. Our answer to why there are inhibitory/negative weights is: to learn more functions. We prove that, in the absence of negative weights, neural networks with non-decreasing activation functions are not universal approximators. While this may be an intuitive result to some, to the best of our knowledge, there is no formal theory, in either machine learning or neuroscience, that demonstrates why negative weights are crucial in the context of representation capacity. Further, we provide insights on the geometric properties of the representation space that non-negative deep networks cannot represent. We expect these insights will yield a deeper understanding of more sophisticated inductive priors imposed on the distribution of weights that lead to more efficient biological and machine learning.

Natural intelligences (NIs) thrive in a dynamic world - they learn quickly, sometimes with only a few samples. In contrast, artificial intelligences (AIs) typically learn with a prohibitive number of training samples and computational power. What design principle difference between NI and AI could contribute to such a discrepancy? Here, we investigate the role of weight polarity: development processes initialize NIs with advantageous polarity configurations; as NIs grow and learn, synapse magnitudes update, yet polarities are largely kept unchanged. We demonstrate with simulation and image classification tasks that if weight polarities are adequately set a priori, then networks learn with less time and data. We also explicitly illustrate situations in which a priori setting the weight polarities is disadvantageous for networks. Our work illustrates the value of weight polarities from the perspective of statistical and computational efficiency during learning.

Generalization error always improves with more in-distribution data. However, it is an open question what happens as we add out-of-distribution (OOD) data. Intuitively, if the OOD data is quite different, it seems more data would harm generalization error, though if the OOD data are sufficiently similar, much empirical evidence suggests that OOD data can actually improve generalization error. We show a counter-intuitive phenomenon: the generalization error of a task can be a non-monotonic function of the amount of OOD data. Specifically, we prove that generalization error can improve with small amounts of OOD data, and then get worse than no OOD data with larger amounts. In other words, there is value in training on small amounts of OOD data. We analytically demonstrate these results via Fisher’s Linear Discriminant on synthetic datasets, and empirically demonstrate them via deep networks on computer vision benchmarks such as MNIST, CIFAR-10, CINIC-10, PACS and DomainNet. In the idealistic setting where we know which samples are OOD, we show that these non-monotonic trends can be exploited using an appropriately weighted objective of the target and OOD empirical risk. While its practical utility is limited, this does suggest that if we can detect OOD samples, then there may be ways to benefit from them. When we do not know which samples are OOD, we show how a number of go-to strategies such as data-augmentation, hyper-parameter optimization and pre-training are not enough to ensure that the target generalization error does not deteriorate with the number of OOD samples in the dataset.

Learning to rank – producing a ranked list of items specific to a query and with respect to a set of supervisory items – is a problem of general interest. The setting we consider is one in which no analytic description of what constitutes a good ranking is available. Instead, we have a collection of representations and supervisory information consisting of a (target item, interesting items set) pair. We demonstrate analytically, in simulation, and in real data examples that learning to rank via combining representations using an integer linear program is effective when the supervision is as light as “these few items are similar to your item of interest.” While this nomination task is quite general, for specificity we present our methodology from the perspective of vertex nomination in graphs. The methodology described herein is model agnostic.

We present methods to serialize and deserialize gradient-boosted trees and random forests that optimize inference latency when models are not loaded into memory. This arises when models are larger than memory, but also systematically when models are deployed on low-resource devices in the Internet of Things or run as cloud microservices where resources are allocated on demand. Block-Aligned Serialized Trees (BLOCKSET) introduce the concept of selective access for random forests and gradient boosted trees in which only the parts of the model needed for inference are deserialized and loaded into memory. %BLOCKSET combines concepts from external memory algorithms and data-parallel %layouts of random forests that maximize I/O-density for in-memory models. Using principles from external memory algorithms, we block-align the serialization format in order to minimize the number of I/Os. For gradient boosted trees, this results in a more than five time reduction in inference latency over layouts that do not perform selective access and a 2 times latency reduction over techniques that are selective, but do not encode I/O block boundaries in the layout.

The development of models and methodology for the analysis of data from multiple heterogeneous networks is of importance both in statistical network theory and across a wide spectrum of application domains. Although single-graph analysis is well-studied, multiple graph inference is largely unexplored, in part because of the challenges inherent in appropriately modeling graph differences and yet retaining sufficient model simplicity to render estimation feasible. This paper addresses exactly this gap, by introducing a new model, the common subspace independent-edge multiple random graph model, which describes a heterogeneous collection of networks with a shared latent structure on the vertices but potentially different connectivity patterns for each graph. The model encompasses many popular network representations, including the stochastic blockmodel. The model is both flexible enough to meaningfully account for important graph differences, and tractable enough to allow for accurate inference in multiple networks. In particular, a joint spectral embedding of adjacency matrices---the multiple adjacency spectral embedding---leads to simultaneous consistent estimation of underlying parameters for each graph. Under mild additional assumptions, the estimates satisfy asymptotic normality and yield improvements for graph eigenvalue estimation. In both simulated and real data, the model and the embedding can be deployed for a number of subsequent network inference tasks, including dimensionality reduction, classification, hypothesis testing, and community detection. Specifically, when the embedding is applied to a data set of connectomes constructed through diffusion magnetic resonance imaging, the result is an accurate classification of brain scans by human subject and a meaningful determination of heterogeneity across scans of different individuals.

As data are generated more and more from multiple disparate sources, multiview data sets, where each sample has features in distinct views, have grown in recent years. However, no comprehensive package exists that enables non-specialists to use these methods easily. mvlearn is a Python library which implements the leading multiview machine learning methods. Its simple API closely follows that of scikit-learn for increased ease-of-use. The package can be installed from Python Package Index (PyPI) and the conda package manager and is released under the MIT open-source license. The documentation, detailed examples, and all releases are available at https://mvlearn.github.io/.

Energy statistics was proposed by Székely in the 80's inspired by Newton's gravitational potential in classical mechanics and it provides a model-free hypothesis test for equality of distributions. In its original form, energy statistics was formulated in euclidean spaces. More recently, it was generalized to metric spaces of negative type. In this paper, we consider a formulation for the clustering problem using a weighted version of energy statistics in spaces of negative type. We show that this approach leads to a quadratically constrained quadratic program in the associated kernel space, establishing connections with graph partitioning problems and kernel methods in machine learning. To find local solutions of such an optimization problem, we propose kernel k-groups, which is an extension of Hartigan's method to kernel spaces. Kernel k-groups is cheaper than spectral clustering and has the same computational cost as kernel k-means (which is based on Lloyd's heuristic) but our numerical results show an improved performance, especially in higher dimensions. Moreover, we verify the efficiency of kernel k-groups in community detection in sparse stochastic block models which has fascinating applications in several areas of science.

Feature extraction and dimension reduction for networks is critical in a wide variety of domains. Efficiently and accurately learning features for multiple graphs has important applications in statistical inference on graphs. We propose a method to jointly embed multiple undirected graphs. Given a set of graphs, the joint embedding method identifies a linear subspace spanned by rank one symmetric matrices and projects adjacency matrices of graphs into this subspace. The projection coefficients can be treated as features of the graphs, while the embedding components can represent vertex features. We also propose a random graph model for multiple graphs that generalizes other classical models for graphs. We show through theory and numerical experiments that under the model, the joint embedding method produces estimates of parameters with small errors. Via simulation experiments, we demonstrate that the joint embedding method produces features which lead to state of the art performance in classifying graphs. Applying the joint embedding method to human brain graphs, we find it extracts interpretable features with good prediction accuracy in different tasks.

Replicability, the ability to replicate scientific findings, is a prerequisite for scientific discovery and clinical utility. Troublingly, we are in the midst of a replicability crisis. A key to replicability is that multiple measurements of the same item (e.g., experimental sample or clinical participant) under fixed experimental constraints are relatively similar to one another. Thus, statistics that quantify the relative contributions of accidental deviations-such as measurement error-as compared to systematic deviations-such as individual differences-are critical. We demonstrate that existing replicability statistics, such as intra-class correlation coefficient and fingerprinting, fail to adequately differentiate between accidental and systematic deviations in very simple settings. We therefore propose a novel statistic, discriminability, which quantifies the degree to which an individual's samples are relatively similar to one another, without restricting the data to be univariate, Gaussian, or even Euclidean. Using this statistic, we introduce the possibility of optimizing experimental design via increasing discriminability and prove that optimizing discriminability improves performance bounds in subsequent inference tasks. In extensive simulated and real datasets (focusing on brain imaging and demonstrating on genomics), only optimizing data discriminability improves performance on all subsequent inference tasks for each dataset. We therefore suggest that designing experiments and analyses to optimize discriminability may be a crucial step in solving the replicability crisis, and more generally, mitigating accidental measurement error.

Together with the curse of dimensionality, nonlinear dependencies in large data sets persist as major challenges in data mining tasks. A reliable way to accurately preserve nonlinear structure is to compute geodesic distances between data points. Manifold learning methods, such as Isomap, aim to preserve geodesic distances in a Riemannian manifold. However, as manifold learning algorithms operate on the ambient dimensionality of the data, the essential step of geodesic distance computation is sensitive to high-dimensional noise. Therefore, a direct application of these algorithms to high-dimensional, noisy data often yields unsatisfactory results and does not accurately capture nonlinear structure.

Decision forests, including Random Forests and Gradient Boosting Trees, have recently demonstrated state-of-the-art performance in a variety of machine learning settings. Decision forests are typically ensembles of axis-aligned decision trees; that is, trees that split only along feature dimensions. In contrast, many recent extensions to decision forests are based on axis-oblique splits. Unfortunately, these extensions forfeit one or more of the favorable properties of decision forests based on axis-aligned splits, such as robustness to many noise dimensions, interpretability, or computational efficiency. We introduce yet another decision forest, called “Sparse Projection Oblique Randomer Forests” (SPORF). SPORF trees recursively split along very sparse random projections. Our method significantly improves accuracy over existing state-of-the-art algorithms on a standard benchmark suite for classification with $>100$ problems of varying dimension, sample size, and number of classes. To illustrate how SPORF addresses the limitations of both axis-aligned and existing oblique decision forest methods, we conduct extensive simulated experiments. SPORF typically yields improved performance over existing decision forest methods, while mitigating computational efficiency and scalability and maintaining interpretability. Very sparse random projections can be incorporated into gradient boosted trees to obtain potentially similar gains.

Decision Forests are popular machine learning techniques that assist scientists to extract knowledge from massive data sets. This class of tool remains popular because of their interpretability and ease of use, unlike other modern machine learning methods, such as kernel machines and deep learning. Decision forests also scale well for use with large data because training and run time operations are trivially parallelizable allowing for high inference throughputs. A negative aspect of these forests, and an untenable property for many real time applications, is their high inference latency caused by the combination of large model sizes with random memory access patterns. We present memory packing techniques and a novel tree traversal method to overcome this deficiency. The result of our system is a grouping of trees into a hierarchical structure. At low levels, we pack the nodes of multiple trees into contiguous memory blocks so that each memory access fetches data for multiple trees. At higher levels, we use leaf cardinality to identify the most popular paths through a tree and collocate those paths in contiguous cache lines. We extend this layout with a re-ordering of the tree traversal algorithm to take advantage of the increased memory throughput provided by out-of-order execution and cache-line prefetching. Together, these optimizations increase the performance and parallel scalability of classification in ensembles by a factor of ten over an optimized C++ implementation and a popular R-language implementation.

We introduce graspy, a Python library devoted to statistical inference, machine learning, and visualization of random graphs and graph populations. This package provides flexible and easy-to-use algorithms for analyzing and understanding graphs with a sklearn compliant API. graspy can be downloaded from Python Package Index (PyPi), and is released under the Apache 2.0 open-source license. The documentation and all releases are available at https://neurodata.io/graspy.

In brain imaging and connectomics, the study of brain networks, estimating the mean of a population of graphs based on a sample is a core problem. Often, this problem is especially difficult because the sample or cohort size is relatively small, sometimes even a single subject, while the number of nodes can be very large with noisy estimates of connectivity. While the element-wise sample mean of the adjacency matrices is a common approach, this method does not exploit the underlying structural properties of the graphs. We propose using a low-rank method that incorporates dimension selection and diagonal augmentation to smooth the estimates and improve performance over the naïve methodology for small sample sizes. Theoretical results for the stochastic block model show that this method offers major improvements when there are many vertices. Similarly, we demonstrate that the low-rank methods outperform the standard sample mean for a variety of independent edge distributions as well as human connectome data derived from the magnetic resonance imaging, especially when the sample sizes are small. Moreover, the low-rank methods yield "eigen-connectomes," which correlate with the lobe-structure of the human brain and superstructures of the mouse brain. These results indicate that the low-rank methods are the important parts of the toolbox for researchers studying populations of graphs in general and statistical connectomics in particular.

We describe NDStore, a scalable multi-hierarchical data storage deployment for spatial analysis of neuroscience data on the AWS cloud. The system design is inspired by the requirement to maintain high I/O throughput for workloads that build neural connectivity maps of the brain from peta-scale imaging data using computer vision algorithms. We store all our data on the AWS object store S3 to limit our deployment costs. S3 serves as our base-tier of storage. Redis, an in-memory key-value engine, is used as our caching tier. The data is dynamically moved between the different storage tiers based on user access. All programming interfaces to this system are RESTful web-services. We include a performance evaluation that shows that our production system provides good performance for a variety of workloads by combining the assets of multiple cloud services.

R is one of the most popular programming languages for statistics and machine learning, but it is slow and unable to scale to large datasets. The general approach for having an efficient algorithm in R is to implement it in C or FORTRAN and provide an R wrapper. FlashR accelerates and scales existing R code by parallelizing a large number of matrix functions in the R base package and scaling them beyond memory capacity with solid-state drives (SSDs). FlashR performs memory hierarchy aware execution to speed up parallelized R code by (i) evaluating matrix operations lazily, (ii) performing all operations in a DAG in a single execution and with only one pass over data to increase the ratio of computation to I/O, (iii) performing two levels of matrix partitioning and reordering computation on matrix partitions to reduce data movement in the memory hierarchy. We evaluate FlashR on various machine learning and statistics algorithms on inputs of up to four billion data points. Despite the huge performance gap between SSDs and RAM, FlashR on SSDs closely tracks the performance of FlashR in memory for many algorithms. The R implementations in FlashR outperforms H2O and Spark MLlib by a factor of 3 -- 20.

k-means is one of the most influential and utilized machine learning algorithms. Its computation limits the performance and scalability of many statistical analysis and machine learning tasks. We rethink and optimize k-means in terms of modern NUMA architectures to develop a novel parallelization scheme that delays and minimizes synchronization barriers. The k-means NUMA Optimized Routine knor) library has (i) in-memory knori), (ii) distributed memory (knord), and (ii) semi-external memory (\textsf{knors}) modules that radically improve the performance of k-means for varying memory and hardware budgets. knori boosts performance for single machine datasets by an order of magnitude or more. \textsf{knors} improves the scalability of k-means on a memory budget using SSDs. knors scales to billions of points on a single machine, using a fraction of the resources that distributed in-memory systems require. knord retains knori's performance characteristics, while scaling in-memory through distributed computation in the cloud. knor modifies Elkan's triangle inequality pruning algorithm such that we utilize it on billion-point datasets without the significant memory overhead of the original algorithm. We demonstrate knor outperforms distributed commercial products like H2O, Turi (formerly Dato, GraphLab) and Spark's MLlib by more than an order of magnitude for datasets of 107 to 109 points.

Random Forest (RF) remains one of the most widely used general purpose classification methods. Two recent large-scale empirical studies demonstrated it to be the best overall classification method among a variety of methods evaluated. One of its main limitations, however, is that it is restricted to only axis-aligned recursive partitions of the feature space. Consequently, RF is particularly sensitive to the orientation of the data. Several studies have proposed “oblique” decision forest methods to address this limitation. However, these methods either have a time and space complexity significantly greater than RF, are sensitive to unit and scale, or empirically do not perform as well as RF on real data. One promising oblique method that was proposed alongside the canonical RF method, called Forest-RC (F-RC), has not received as much attention by the community. Despite it being just as old as RF, virtually no studies exist investigating its theoretical or empirical performance. In this work, we demonstrate that F-RC empirically outperforms RF and another recently proposed oblique method called Random Rotation Random Forest, while approximately maintaining the same computational complexity. Furthermore, a variant of F-RC which rank transforms the data prior to learning is especially invariant to affine transformations and robust to data corruption. Open source code is available.

Deeper exploration of the brain's vast synaptic networks will require new tools for high-throughput structural and molecular profiling of the diverse populations of synapses that compose those networks. Fluorescence microscopy (FM) and electron microscopy (EM) offer complementary advantages and disadvantages for single-synapse analysis. FM combines exquisite molecular discrimination capacities with high speed and low cost, but rigorous discrimination between synaptic and non-synaptic fluorescence signals is challenging. In contrast, EM remains the gold standard for reliable identification of a synapse, but offers only limited molecular discrimination and is slow and costly. To develop and test single-synapse image analysis methods, we have used datasets from conjugate array tomography (cAT), which provides voxel-conjugate FM and EM (annotated) images of the same individual synapses. We report a novel unsupervised probabilistic method for detection of synapses from multiplex FM (muxFM) image data, and evaluate this method both by comparison to EM gold standard annotated data and by examining its capacity to reproduce known important features of cortical synapse distributions. The proposed probabilistic model-based synapse detector accepts molecular-morphological synapse models as user queries, and delivers a volumetric map of the probability that each voxel represents part of a synapse. Taking human annotation of cAT EM data as ground truth, we show that our algorithm detects synapses from muxFM data alone as successfully as human annotators seeing only the muxFM data, and accurately reproduces known architectural features of cortical synapse distributions. This approach opens the door to data-driven discovery of new synapse types and their density. We suggest that our probabilistic synapse detector will also be useful for analysis of standard confocal and super-resolution FM images, where EM cross-validation is not practical.

High-dimensional time-series data from a wide variety of domains, such as neuroscience, are being generated every day. Fitting statistical models to such data, to enable parameter estimation and time-series prediction, is an important computational primitive. Existing methods, however, are unable to cope with the high-dimensional nature of these data, due to both computational and statistical reasons. We mitigate both kinds of issues by proposing an M-estimator for Reduced-rank System IDentification ( Mr. Sid ). A combination of low-rank approximations, ℓ 1 and ℓ 2 penalties, and some numerical linear algebra tricks, yields an estimator that is computationally efficient and numerically stable. Simulations and real data examples demonstrate the usefulness of this approach in a variety of problems. In particular, we demonstrate that Mr. Sid can accurately estimate spatial filters, connectivity graphs, and time-courses from native resolution functional magnetic resonance imaging data. Mr. Sid therefore enables big time-series data to be analyzed using standard methods, readying the field for further generalizations including nonlinear and non-Gaussian state-space models.

Sparse matrix multiplication is traditionally performed in memory and scales to large matrices using the distributed memory of multiple nodes. In contrast, we scale sparse matrix multiplication beyond memory capacity by implementing sparse matrix dense matrix multiplication (SpMM) in a semi-external memory (SEM) fashion; i.e., we keep the sparse matrix on commodity SSDs and dense matrices in memory. Our SEM-SpMM incorporates many in-memory optimizations for large power-law graphs. It outperforms the in-memory implementations of Trilinos and Intel MKL and scales to billion-node graphs, far beyond the limitations of memory. Furthermore, on a single large parallel machine, our SEM-SpMM operates as fast as the distributed implementations of Trilinos using five times as much processing power. We also run our implementation in memory (IM-SpMM) to quantify the overhead of keeping data on SSDs. SEM-SpMM achieves almost 100 percent performance of IM-SpMM on graphs when the dense matrix has more than four columns; it achieves at least 65 percent performance of IM-SpMM on all inputs. We apply our SpMM to three important data analysis tasks-PageRank, eigensolving, and non-negative matrix factorization-and show that our SEM implementations significantly advance the state of the art.

For random graphs distributed according to stochastic blockmodels, a special case of latent position graphs, adjacency spectral embedding followed by appropriate vertex classification is asymptotically Bayes optimal; but this approach requires knowledge of and critically depends on the model dimension. In this paper, we propose a sparse representation vertex classifier which does not require information about the model dimension. This classifier represents a test vertex as a sparse combination of the vertices in the training set and uses the recovered coefficients to classify the test vertex. We prove consistency of our proposed classifier for stochastic blockmodels, and demonstrate that the sparse representation classifier can predict vertex labels with higher accuracy than adjacency spectral embedding approaches via both simulation studies and real data experiments. Our results demonstrate the robustness and effectiveness of our proposed vertex classifier when the model dimension is unknown.

Graph matching-aligning a pair of graphs to minimize their edge disagreements-has received wide-spread attention from both theoretical and applied communities over the past several decades, including combinatorics, computer vision, and connectomics. Its attention can be partially attributed to its computational difficulty. Although many heuristics have previously been proposed in the literature to approximately solve graph matching, very few have any theoretical support for their performance. A common technique is to relax the discrete problem to a continuous problem, therefore enabling practitioners to bring gradient-descent-type algorithms to bear. We prove that an indefinite relaxation (when solved exactly) almost always discovers the optimal permutation, while a common convex relaxation almost always fails to discover the optimal permutation. These theoretical results suggest that initializing the indefinite algorithm with the convex optimum might yield improved practical performance. Indeed, experimental results illuminate and corroborate these theoretical findings, demonstrating that excellent results are achieved in both benchmark and real data problems by amalgamating the two approaches.

How much has a network changed since yesterday? How different is the wiring of Bob’s brain (a left-handed male) and Alice’s brain (a right-handed female), and how is it different? Graph similarity with given node correspondence, i.e., the detection of changes in the connectivity of graphs, arises in numerous settings. In this work, we formally state the axioms and desired properties of the graph similarity functions, and evaluate when state-of-the-art methods fail to detect crucial connectivity changes in graphs. We propose D elta C on , a principled, intuitive, and scalable algorithm that assesses the similarity between two graphs on the same nodes (e.g., employees of a company, customers of a mobile carrier). In conjunction, we propose D elta C on -A ttr , a related approach that enables attribution of change or dissimilarity to responsible nodes and edges. Experiments on various synthetic and real graphs showcase the advantages of our method over existing similarity measures. Finally, we employ D elta C on and D elta C on -A ttr on real applications: (a) we classify people to groups of high and low creativity based on their brain connectivity graphs, (b) do temporal anomaly detection in the who-emails-whom Enron graph and find the top culprits for the changes in the temporal corporate email graph, and (c) recover pairs of test-retest large brain scans ( ∼17M edges, up to 90M edges) for 21 subjects.

With simultaneous measurements from ever increasing populations of neurons, there is a growing need for sophisticated tools to recover signals from individual neurons. In electrophysiology experiments, this classically proceeds in a two-step process: (i) threshold the waveforms to detect putative spikes and (ii) cluster the waveforms into single units (neurons). We extend previous Bayesian nonparamet- ric models of neural spiking to jointly detect and cluster neurons using a Gamma process model. Importantly, we develop an online approximate inference scheme enabling real-time analysis, with performance exceeding the previous state-of-the- art. Via exploratory data analysis—using data with partial ground truth as well as two novel data sets—we find several features of our model collectively contribute to our improved performance including: (i) accounting for colored noise, (ii) de- tecting overlapping spikes, (iii) tracking waveform dynamics, and (iv) using mul- tiple channels. We hope to enable novel experiments simultaneously measuring many thousands of neurons and possibly adapting stimuli dynamically to probe ever deeper into the mysteries of the brain.

Graph matching is a challenging problem with very important applications in a wide range of fields, from image and video analysis to biological and biomedical problems. We propose a robust graph matching algorithm inspired in sparsity-related techniques. We cast the problem, resembling group or collaborative sparsity formulations, as a non-smooth convex optimization problem that can be efficiently solved using augmented Lagrangian techniques. The method can deal with weighted or unweighted graphs, as well as multimodal data, where different graphs represent different types of data. The proposed approach is also naturally integrated with collaborative graph inference techniques, solving general network inference problems where the observed variables, possibly coming from different modalities, are not in correspondence. The algorithm is tested and compared with state-of-the-art graph matching techniques in both synthetic and real graphs. We also present results on multimodal graphs and applications to collaborative inference of brain connectivity from alignment-free functional magnetic resonance imaging (fMRI) data.

Nonparametric estimation of the conditional distribution of a response given high-dimensional features is a challenging problem. It is important to allow not only the mean but also the variance and shape of the response density to change flexibly with features, which are massive-dimensional. We propose a multiscale dictionary learning model, which expresses the conditional response density as a convex combination of dictionary densities, with the densities used and their weights dependent on the path through a tree decomposition of the feature space. A fast graph partitioning algorithm is applied to obtain the tree decomposition, with Bayesian methods then used to adaptively prune and average over different sub-trees in a soft probabilistic manner. The algorithm scales efficiently to approximately one million features. State of the art predictive performance is demonstrated for toy examples and two neuroscience applications including up to a million features.

How much did a network change since yesterday? How different is the wiring between Bob's brain (a left-handed male) and Alice's brain (a right-handed female)? Graph similarity with known node correspondence, i.e. the detection of changes in the connectivity of graphs, arises in numerous settings. In this work, we formally state the axioms and desired properties of the graph similarity functions, and evaluate when state-of-the-art methods fail to detect crucial connectivity changes in graphs. We propose DeltaCon, a principled, intuitive, and scalable algorithm that assesses the similarity between two graphs on the same nodes (e.g. employees of a company, customers of a mobile carrier). Experiments on various synthetic and real graphs showcase the advantages of our method over existing similarity measures. Finally, we employ DeltaCon to real applications: (a) we classify people to groups of high and low creativity based on their brain connectivity graphs, and (b) do temporal anomaly detection in the who-emails-whom Enron graph.

- neural connectivity maps of the brain-using the parallel execution of computer vision algorithms on high-performance compute clusters. These services and open-science data sets are publicly available at openconnecto.me. The system design inherits much from NoSQL scale-out and data-intensive computing architectures. We distribute data to cluster nodes by partitioning a spatial index. We direct I/O to different systems-reads to parallel disk arrays and writes to solid-state storage-to avoid I/O interference and maximize throughput. All programming interfaces are RESTful Web services, which are simple and stateless, improving scalability and usability. We include a performance evaluation of the production system, highlighting the effec-tiveness of spatial data organization.

This manuscript considers the following "graph classification" question: Given a collection of graphs and associated classes, how can one predict the class of a newly observed graph? To address this question, we propose a statistical model for graph/class pairs. This model naturally leads to a set of estimators to identify the class-conditional signal, or "signal-subgraph," defined as the collection of edges that are probabilistically different between the classes. The estimators admit classifiers which are asymptotically optimal and efficient, but which differ by their assumption about the "coherency" of the signal-subgraph (coherency is the extent to which the signal-edges "stick together" around a common subset of vertices). Via simulation, the best estimator is shown to be not just a function of the coherency of the model, but also the number of training samples. These estimators are employed to address a contemporary neuroscience question: Can we classify "connectomes" (brain-graphs) according to sex? The answer is yes, and significantly better than all benchmark algorithms considered. Synthetic data analysis demonstrates that even when the model is correct, given the relatively small number of training samples, the estimated signal-subgraph should be taken with a grain of salt. We conclude by discussing several possible extensions.

Decision making lies at the very heart of many psychiatric diseases. It is also a central theoretical concern in a wide variety of fields and has undergone detailed, in-depth, analyses. We take as an example Major Depressive Disorder (MDD), applying insights from a Bayesian reinforcement learning framework. We focus on anhedonia and helplessness. Helplessness—a core element in the conceptual- izations of MDD that has lead to major advances in its treatment, pharmacolog- ical and neurobiological understanding—is formalized as a simple prior over the outcome entropy of actions in uncertain environments. Anhedonia, which is an equally fundamental aspect of the disease, is related to the effective reward size. These formulations allow for the design of specific tasks to measure anhedonia and helplessness behaviorally. We show that these behavioral measures capture explicit, questionnaire-based cognitions. We also provide evidence that these tasks may allow classification of subjects into healthy and MDD groups based purely on a behavioural measure and avoiding any verbal reports. There are strong ties between decision making and psychiatry, with maladaptive decisions and be- haviors being very prominent in people with psychiatric disorders. Depression is classically seen as following life events such as divorces and job losses. Longitudinal studies, however, have revealed that a significant fraction of the stressors associated with depression do in fact follow MDD onset, and that they are likely due to maladaptive behaviors prominent in MDD (Kendler et al., 1999). Clinically effective ’talking’ therapies for MDD such as cognitive and dialectical behavior therapies (DeRubeis et al., 1999; Bortolotti et al., 2008; Gotlib and Hammen, 2002; Power, 2005) explicitly concentrate on altering patients’ maladaptive behaviors and decision making processes. Decision making is a promising avenue into psychiatry for at least two more reasons. First, it offers powerful analytical tools. Control problems related to decision making are prevalent in a huge diversity of fields, ranging from ecology to economics, computer science and engineering. These fields have produced well-founded and thoroughly characterized frameworks within which many issues in decision making can be framed. Here, we will focus on framing issues identified in psychiatric settings within a normative decision making framework. Its second major strength comes from its relationship to neurobiology, and particularly those neuro- modulatory systems which are powerfully affected by all major clinically effective pharmacothera- pies in psychiatry. The understanding of these systems has benefited significantly from theoretical accounts of optimal control such as reinforcement learning (Montague et al., 1996; Kapur and Rem- ington, 1996; Smith et al., 1999; Yu and Dayan, 2005; Dayan and Yu, 2006). Such accounts may be useful to identify in more specific terms the roles of the neuromodulators in psychiatry (Smith et al., 2004; Williams and Dayan, 2005; Moutoussis et al., 2008; Dayan and Huys, 2008). ∗[email protected], [email protected], [email protected]; www.gatsby.ucl.ac.uk/∼qhuys/pub.html

A mixed-signal very large scale integration (VLSI) chip for large scale emulation of spiking neural networks is presented. The chip contains 2400 silicon neurons with fully programmable and reconfigurable synaptic connectivity. Each neuron implements a discrete-time model of a single-compartment cell. The model allows for analog membrane dynamics and an arbitrary number of synaptic connections, each with tunable conductance and reversal potential. The array of silicon neurons functions as an address-event (AE) transceiver, with incoming and outgoing spikes communicated over an asynchronous event-driven digital bus. Address encoding and conflict resolution of spiking events are implemented via a randomized arbitration scheme that ensures balanced servicing of event requests across the array. Routing of events is implemented externally using dynamically programmable random-access memory that stores a postsynaptic address, the conductance, and the reversal potential of each synaptic connection. Here, we describe the silicon neuron circuits, present experimental data characterizing the 3 mm x 3 mm chip fabricated in 0.5-microm complementary metal-oxide-semiconductor (CMOS) technology, and demonstrate its utility by configuring the hardware to emulate a model of attractor dynamics and waves of neural activity during sleep in rat hippocampus.