arXiv subjects

Tobias Glasmachers

Publications and source records attributed to Tobias Glasmachers.

At least 55 records · Page 3Linked to original sources

arXiv2019

SpikeDeep-Classifier: A deep-learning based fully automatic offline spike sorting algorithm

Objective. Recent advancements in electrode designs and micro-fabrication technology has allowed existence of microelectrode arrays with hundreds of channels for single-cell recordings. In such electrophysiological recordings, each implanted micro-electrode can record the activities of more than one neuron in its vicinity. Recording the activities of multiple neurons may also be referred to as multiple unit activity. However, for any further analysis, the main goal is to isolate the activity of each recorded neuron and thus called single-unit activity. This process may also be referred to as spike sorting or spike classification. Recent approaches to extract SUA are time consuming, mainly due to the requirement of human intervention at various stages of spike sorting pipeline. Lack of standardization is another drawback of the current available approaches. Therefore, in this study we proposed a standard spike sorter: SpikeDeep-Classifier, a fully automatic spike sorting algorithm. Approach. We proposed a novel spike sorting pipeline, based on a set of supervised and unsupervised learning algorithms. We used supervised, deep learning-based algorithms for extracting meaningful channels and removing background activities (noise) from the extracted channels. We also showed that the process of clustering becomes straight-forward, once the noise/artifact is completely removed from the data. Therefore, in the next stage, we applied a simple clustering algorithm (K-mean) with predefined maximum number of clusters. Lastly, we used a similarity-based criterion to keep distinct clusters and merge similar-looking clusters. Main results. We evaluated our algorithm on a dataset collected from two different species (humans and non-human primates (NHPs)) without any retraining. We also validated our algorithm on two publicly available labeled datasets.

q-bio.QM↗

arXiv2019

Vehicle Shape and Color Classification Using Convolutional Neural Network

This paper presents a module of vehicle reidentification based on make/model and color classification. It could be used by the Automated Vehicular Surveillance (AVS) or by the fast analysis of video data. Many of problems, that are related to this topic, had to be addressed. In order to facilitate and accelerate the progress in this subject, we will present our way to collect and to label a large scale data set. We used deeper neural networks in our training. They showed a good classification accuracy. We show the results of make/model and color classification on controlled and video data set. We demonstrate with the help of a developed application the re-identification of vehicles on video images based on make/model and color classification. This work was partially funded under the grant.

cs.CV↗

arXiv2019

Challenges of Convex Quadratic Bi-objective Benchmark Problems

Convex quadratic objective functions are an important base case in state-of-the-art benchmark collections for single-objective optimization on continuous domains. Although often considered rather simple, they represent the highly relevant challenges of non-separability and ill-conditioning. In the multi-objective case, quadratic benchmark problems are under-represented. In this paper we analyze the specific challenges that can be posed by quadratic functions in the bi-objective case. Our construction yields a full factorial design of 54 different problem classes. We perform experiments with well-established algorithms to demonstrate the insights that can be supported by this function class. We find huge performance differences, which can be clearly attributed to two root causes: non-separability and alignment of the Pareto set with the coordinate system.

cs.NE↗

arXiv2019

Drift Theory in Continuous Search Spaces: Expected Hitting Time of the (1+1)-ES with 1/5 Success Rule

This paper explores the use of the standard approach for proving runtime bounds in discrete domains---often referred to as drift analysis---in the context of optimization on a continuous domain. Using this framework we analyze the (1+1) Evolution Strategy with one-fifth success rule on the sphere function. To deal with potential functions that are not lower-bounded, we formulate novel drift theorems. We then use the theorems to prove bounds on the expected hitting time to reach a certain target fitness in finite dimension $d$. The bounds are akin to linear convergence. We then study the dependency of the different terms on $d$ proving a convergence rate dependency of $Θ(1/d)$. Our results constitute the first non-asymptotic analysis for the algorithm considered as well as the first explicit application of drift analysis to a randomized search heuristic with continuous domain.

cs.NE↗

arXiv2018

Universal Spike Classifier

In electrophysiology, microelectrodes are the primary source for recording neural data of single neurons (single unit activity). These microelectrodes can be implanted individually, or in the form of microelectrodes arrays, consisting of hundreds of electrodes. During recordings, some channels capture the activity of neurons, which is usually contaminated with external artifacts and noise. Another considerable fraction of channels does not record any neural data, but external artifacts and noise. Therefore, an automatic identification and tracking of channels containing neural data is of great significance and can accelerate the process of analysis, e.g. automatic selection of meaningful channels during offline and online spike sorting. Another important aspect is the selection of meaningful channels during online decoding in brain-computer interface applications, where threshold crossing events are usually for feature extraction, even though they do not necessarily correspond to neural events. Here, we propose a novel algorithm based on the newly introduced way of feature vector extraction and a supervised deep learning method: a universal spike classifier (USC). The USC enables us to address both above-raised issues. The USC uses the standard architecture of convolutional neural networks (Conv net). It takes the batch of the waveforms, instead of a single waveform as an input, propagates it through the multilayered structure, and finally classifies it as a channel containing neural spike data or artifacts. We have trained the model of USC on data recorded from single tetraplegic patient with Utah arrays implanted in different brain areas. This trained model was then evaluated without retraining on the data collected from six epileptic patients implanted with depth electrodes and two tetraplegic patients implanted with two Utah arrays, individually.

q-bio.NC↗

arXiv2018

Challenges in High-dimensional Reinforcement Learning with Evolution Strategies

Evolution Strategies (ESs) have recently become popular for training deep neural networks, in particular on reinforcement learning tasks, a special form of controller design. Compared to classic problems in continuous direct search, deep networks pose extremely high-dimensional optimization problems, with many thousands or even millions of variables. In addition, many control problems give rise to a stochastic fitness function. Considering the relevance of the application, we study the suitability of evolution strategies for high-dimensional, stochastic problems. Our results give insights into which algorithmic mechanisms of modern ES are of value for the class of problems at hand, and they reveal principled limitations of the approach. They are in line with our theoretical understanding of ESs. We show that combining ESs that offer reduced internal algorithm cost with uncertainty handling techniques yields promising methods for this class of problems.

cs.NE↗

arXiv2018

Multi-Merge Budget Maintenance for Stochastic Gradient Descent SVM Training

Budgeted Stochastic Gradient Descent (BSGD) is a state-of-the-art technique for training large-scale kernelized support vector machines. The budget constraint is maintained incrementally by merging two points whenever the pre-defined budget is exceeded. The process of finding suitable merge partners is costly; it can account for up to 45% of the total training time. In this paper we investigate computationally more efficient schemes that merge more than two points at once. We obtain significant speed-ups without sacrificing accuracy.

cs.LG↗

arXiv2018

Speeding Up Budgeted Stochastic Gradient Descent SVM Training with Precomputed Golden Section Search

Limiting the model size of a kernel support vector machine to a pre-defined budget is a well-established technique that allows to scale SVM learning and prediction to large-scale data. Its core addition to simple stochastic gradient training is budget maintenance through merging of support vectors. This requires solving an inner optimization problem with an iterative method many times per gradient step. In this paper we replace the iterative procedure with a fast lookup. We manage to reduce the merging time by up to 65% and the total training time by 44% without any loss of accuracy.

cs.LG↗

arXiv2018

Dual SVM Training on a Budget

We present a dual subspace ascent algorithm for support vector machine training that respects a budget constraint limiting the number of support vectors. Budget methods are effective for reducing the training time of kernel SVM while retaining high accuracy. To date, budget training is available only for primal (SGD-based) solvers. Dual subspace ascent methods like sequential minimal optimization are attractive for their good adaptation to the problem structure, their fast convergence rate, and their practical speed. By incorporating a budget constraint into a dual algorithm, our method enjoys the best of both worlds. We demonstrate considerable speed-ups over primal budget training methods.

cs.LG↗

arXiv2017

Limited-Memory Matrix Adaptation for Large Scale Black-box Optimization

The Covariance Matrix Adaptation Evolution Strategy (CMA-ES) is a popular method to deal with nonconvex and/or stochastic optimization problems when the gradient information is not available. Being based on the CMA-ES, the recently proposed Matrix Adaptation Evolution Strategy (MA-ES) provides a rather surprising result that the covariance matrix and all associated operations (e.g., potentially unstable eigendecomposition) can be replaced in the CMA-ES by a updated transformation matrix without any loss of performance. In order to further simplify MA-ES and reduce its $\mathcal{O}\big(n^2\big)$ time and storage complexity to $\mathcal{O}\big(n\log(n)\big)$, we present the Limited-Memory Matrix Adaptation Evolution Strategy (LM-MA-ES) for efficient zeroth order large-scale optimization. The algorithm demonstrates state-of-the-art performance on a set of established large-scale benchmarks. We explore the algorithm on the problem of generating adversarial inputs for a (non-smooth) random forest classifier, demonstrating a surprising vulnerability of the classifier.

cs.NE↗

arXiv2017

Limits of End-to-End Learning

End-to-end learning refers to training a possibly complex learning system by applying gradient-based learning to the system as a whole. End-to-end learning system is specifically designed so that all modules are differentiable. In effect, not only a central learning machine, but also all "peripheral" modules like representation learning and memory formation are covered by a holistic learning process. The power of end-to-end learning has been demonstrated on many tasks, like playing a whole array of Atari video games with a single architecture. While pushing for solutions to more challenging tasks, network architectures keep growing more and more complex. In this paper we ask the question whether and to what extent end-to-end learning is a future-proof technique in the sense of scaling to complex and diverse data processing architectures. We point out potential inefficiencies, and we argue in particular that end-to-end learning does not make optimal use of the modular design of present neural networks. Our surprisingly simple experiments demonstrate these inefficiencies, up to the complete breakdown of learning.

cs.LG↗

arXiv2016

A Fast Incremental BSP Tree Archive for Non-dominated Points

Maintaining an archive of all non-dominated points is a standard task in multi-objective optimization. Sometimes it is sufficient to store all evaluated points and to obtain the non-dominated subset in a post-processing step. Alternatively the non-dominated set can be updated on the fly. While keeping track of many non-dominated points efficiently is easy for two objectives, we propose an efficient algorithm based on a binary space partitioning (BSP) tree for the general case of three or more objectives. Our analysis and our empirical results demonstrate the superiority of the method over the brute-force baseline method, as well as graceful scaling to large numbers of objectives.

cs.DS↗

arXiv2016

Anytime Bi-Objective Optimization with a Hybrid Multi-Objective CMA-ES (HMO-CMA-ES)

We propose a multi-objective optimization algorithm aimed at achieving good anytime performance over a wide range of problems. Performance is assessed in terms of the hypervolume metric. The algorithm called HMO-CMA-ES represents a hybrid of several old and new variants of CMA-ES, complemented by BOBYQA as a warm start. We benchmark HMO-CMA-ES on the recently introduced bi-objective problem suite of the COCO framework (COmparing Continuous Optimizers), consisting of 55 scalable continuous optimization problems, which is used by the Black-Box Optimization Benchmarking (BBOB) Workshop 2016.

cs.NE↗

arXiv2016

Fast model selection by limiting SVM training times

Kernelized Support Vector Machines (SVMs) are among the best performing supervised learning methods. But for optimal predictive performance, time-consuming parameter tuning is crucial, which impedes application. To tackle this problem, the classic model selection procedure based on grid-search and cross-validation was refined, e.g. by data subsampling and direct search heuristics. Here we focus on a different aspect, the stopping criterion for SVM training. We show that by limiting the training time given to the SVM solver during parameter tuning we can reduce model selection times by an order of magnitude.

stat.ML↗

arXiv2014

Coordinate Descent with Online Adaptation of Coordinate Frequencies

Coordinate descent (CD) algorithms have become the method of choice for solving a number of optimization problems in machine learning. They are particularly popular for training linear models, including linear support vector machine classification, LASSO regression, and logistic regression. We consider general CD with non-uniform selection of coordinates. Instead of fixing selection frequencies beforehand we propose an online adaptation mechanism for this important parameter, called the adaptive coordinate frequencies (ACF) method. This mechanism removes the need to estimate optimal coordinate frequencies beforehand, and it automatically reacts to changing requirements during an optimization run. We demonstrate the usefulness of our ACF-CD approach for a variety of optimization problems arising in machine learning contexts. Our algorithm offers significant speed-ups over state-of-the-art training methods.

stat.ML↗

arXiv2013

The Planning-ahead SMO Algorithm

The sequential minimal optimization (SMO) algorithm and variants thereof are the de facto standard method for solving large quadratic programs for support vector machine (SVM) training. In this paper we propose a simple yet powerful modification. The main emphasis is on an algorithm improving the SMO step size by planning-ahead. The theoretical analysis ensures its convergence to the optimum. Experiments involving a large number of datasets were carried out to demonstrate the superiority of the new algorithm.

cs.LG↗

arXiv2013

Testing Hypotheses by Regularized Maximum Mean Discrepancy

Do two data samples come from different distributions? Recent studies of this fundamental problem focused on embedding probability distributions into sufficiently rich characteristic Reproducing Kernel Hilbert Spaces (RKHSs), to compare distributions by the distance between their embeddings. We show that Regularized Maximum Mean Discrepancy (RMMD), our novel measure for kernel-based hypothesis testing, yields substantial improvements even when sample sizes are small, and excels at hypothesis tests involving multiple comparisons with power control. We derive asymptotic distributions under the null and alternative hypotheses, and assess power control. Outstanding results are obtained on: challenging EEG data, MNIST, the Berkley Covertype, and the Flare-Solar dataset.

cs.LG↗

arXiv2013

Accelerated Linear SVM Training with Adaptive Variable Selection Frequencies

Support vector machine (SVM) training is an active research area since the dawn of the method. In recent years there has been increasing interest in specialized solvers for the important case of linear models. The algorithm presented by Hsieh et al., probably best known under the name of the "liblinear" implementation, marks a major breakthrough. The method is analog to established dual decomposition algorithms for training of non-linear SVMs, but with greatly reduced computational complexity per update step. This comes at the cost of not keeping track of the gradient of the objective any more, which excludes the application of highly developed working set selection algorithms. We present an algorithmic improvement to this method. We replace uniform working set selection with an online adaptation of selection frequencies. The adaptation criterion is inspired by modern second order working set selection methods. The same mechanism replaces the shrinking heuristic. This novel technique speeds up training in some cases by more than an order of magnitude.

stat.ML↗