Neuronal computations depend on synaptic connectivity and intrinsic electrophysiological properties. Synaptic connectivity determines which inputs from presynaptic neurons are integrated, while cellular properties determine how inputs are filtered over time. Unlike their biological counterparts, most computational approaches to learning in simulated neural networks are limited to changes in synaptic connectivity. However, if intrinsic parameters change, neural computations are altered drastically. Here, we include the parameters that determine the intrinsic properties, e.g., time constants and reset potential, into the learning paradigm. Using sparse feedback signals that indicate target spike times, and gradient-based parameter updates, we show that the intrinsic parameters can be learned along with the synaptic weights to produce specific input-output functions. Specifically, we use a teacher-student paradigm in which a randomly initialised leaky integrate-and-fire or resonate-and-fire neuron must recover the parameters of a teacher neuron. We show that complex temporal functions can be learned online and without backpropagation through time, relying on event-based updates only. Our results are a step towards online learning of neural computations from ungraded and unsigned sparse feedback signals with a biologically inspired learning mechanism.

The task of natural language parsing can naturally be embedded in the maximum-margin framework for structured output prediction using an appropriate joint feature map and a suitable structured loss function. While there are efficient learning algorithms based on the cutting-plane method for optimizing the resulting quadratic objective with potentially exponential number of linear constraints, their efficiency crucially depends on the inference algorithms used to infer the most violated constraint in a current iteration. In this paper, we derive an extension of the well-known Cocke-Kasami-Younger (CKY) algorithm used for parsing with probabilistic context-free grammars for the case of loss-augmented inference enabling an effective training in the cutting-plane approach. The resulting algorithm is guaranteed to find an optimal solution in polynomial time exceeding the running time of the CKY algorithm by a term, which only depends on the number of possible loss values. In order to demonstrate the feasibility of the presented algorithm, we perform a set of experiments for parsing English sentences.

We introduce a new anomaly detection methodology for data with latent dependency structure. As a particular instantiation, we derive a hidden Markov anomaly detector that extends the regular one-class support vector machine. We optimize the approach, which is non-convex, via a DC (difference of convex functions) algorithm, and show that the parameter v can be conveniently used to control the number of outliers in the model. The empirical evaluation on artificial and real data from the domains of computational biology and computational sustainability shows that the approach can achieve significantly higher anomaly detection performance than the regular one-class SVM.

In many real-world applications, the simplified assumption of independent and identically distributed noise breaks down, and labels can have structured, systematic noise. For example, in brain-computer interface applications, training data is often the result of lengthy experimental sessions, where the attention levels of participants can change over the course of the experiment. In such application cases, structured label noise will cause problems because most machine learning methods assume independent and identically distributed label noise. In this paper, we present a novel methodology for learning and evaluation in presence of systematic label noise. The core of which is a novel extension of support vector data description / one-class SVM that can incorporate latent variables. Controlled simulations on synthetic data and a real-world EEG experiment with 20 subjects from the domain of brain-computer-interfacing show that our method achieves accuracies that go beyond the state of the art.

Deep networks can potentially express a learning problem more efficiently than local learning machines. While deep networks outperform local learning machines on some problems, it is still unclear how their nice representation emerges from their complex structure. We present an analysis based on Gaussian kernels that measures how the representation of the learning problem evolves layer after layer as the deep network builds higher-level abstract representations of the input. We use this analysis to show empirically that deep networks build progressively better representations of the learning problem and that the best representations are obtained when the deep network discriminates only in the last layers.

Kernelizing partial least squares (PLS), an algorithm which has been particularly popular in chemometrics, leads to kernel PLS which has several interesting properties, including a sub-cubic runtime for learning, and an iterative construction of directions which are relevant for predicting the outputs. We show that the kernelization of PLS introduces interesting properties not found in ordinary PLS, giving novel insights into the workings of kernel PLS and the connections to kernel ridge regression and conjugate gradient descent methods. Furthermore, we show how to correctly define the degrees of freedom for kernel PLS and how to efficiently compute an unbiased estimate. Finally, we address the practical problem of model selection. We demonstrate how to use the degrees of freedom estimate to perform effective model selection, and discuss how to implement crossvalidation schemes efficiently.

We show that the relevant information about a classification problem in feature space is contained up to negligible error in a finite number of leading kernel PCA components if the kernel matches the underlying learning problem. Thus, kernels not only transform data sets such that good generalization can be achieved even by linear discriminant functions, but this transformation is also performed in a manner which makes economic use of feature space dimensions. In the best case, kernels provide efficient implicit representations of the data to perform classification. Practically, we propose an algorithm which enables us to recover the subspace and dimensionality relevant for good classification. Our algorithm can therefore be applied (1) to analyze the interplay of data set and kernel in a geometric fashion, (2) to help in model selection, and to (3) de-noise in feature space in order to yield better classification results.

The eigenvalues of the kernel matrix play an important role in a number of kernel methods, in particular, in kernel principal component analysis. It is well known that the eigenvalues of the kernel matrix converge as the number of samples tends to infinity. We derive probabilistic finite sample size bounds on the approximation error of individual eigenvalues which have the important property that the bounds scale with the eigenvalue under consideration, reflecting the actual behavior of the approximation errors as predicted by asymptotic results and observed in numerical simulations. Such scaling bounds have so far only been known for tail sums of eigenvalues. Asymptotically, the bounds presented here have a slower than stochastic rate, but the number of sample points necessary to make this disadvantage noticeable is often unrealistically large. Therefore, under practical conditions, and for all but the largest few eigenvalues, the bounds presented here form a significant improvement over existing non-scaling bounds.

Data clustering describes a set of frequently employed techniques in exploratory data analysis to extract "natural" group structure in data. Such groupings need to be validated to separate the signal in the data from spurious structure. In this context, finding an appropriate number of clusters is a particularly important model selection question. We introduce a measure of cluster stability to assess the validity of a cluster model. This stability measure quantifies the reproducibility of clustering solutions on a second sample, and it can be interpreted as a classification risk with regard to class labels produced by a clustering algorithm. The preferred number of clusters is determined by minimizing this classification risk as a function of the number of clusters. Convincing results are achieved on simulated as well as gene expression data sets. Comparisons to other methods demonstrate the competitive performance of our method and its suitability as a general validation tool for clustering solutions in real-world problems.

Model selection is linked to model assessment, which is the problem of comparing different models, or model parameters, for a specific learning task. For supervised learning, the standard practical technique is cross- validation, which is not applicable for semi-supervised and unsupervised settings. In this paper, a new model assessment scheme is introduced which is based on a notion of stability. The stability measure yields an upper bound to cross-validation in the supervised case, but extends to semi-supervised and unsupervised problems. In the experimental part, the performance of the stability measure is studied for model order se- lection in comparison to standard techniques in this area.

We consider noisy Euclidean traveling salesman problems in the plane, which are random combinatorial problems with underlying structure. Gibbs sampling is used to compute average trajectories, which estimate the underlying structure common to all instances. This procedure requires identifying the exact relationship between permutations and tours. In a learning setting, the average trajec(cid:173) tory is used as a model to construct solutions to new instances sampled from the same source. Experimental results show that the average trajectory can in fact estimate the underlying structure and that overfitting effects occur if the trajectory adapts too closely to a single instance.

A tuning method with reinforcement learning (RL) for the retina encoder (RE) of a retina implant (RI) as a visual prosthesis for blind subjects with retinal degenerations is proposed. The RE simulates retinal information processing in real time by means of spatio-temporal receptive field (RF) filters and generates electrical signals for stimulation of several hundreds of ganglion cells (GC) to regain a modest amount of vision. For each contacted GC, the RE has to be optimized with regard to the patient's perception. The patient's (for the present simulated) evaluative feedback is applied here in a dialog module as a reinforcement signal to train several RL agents in a neural network learning process.