Exploiting label dependencies for improved sample complexity

Multi-label classification exhibits several challenges not present in the binary case. The labels may be interdependent, so that the presence of a certain label affects the probability of other labels’ presence. Thus, exploiting dependencies among the labels could be beneficial for the classifier’s predictive performance. Surprisingly, only a few of the existing algorithms address this issue directly by identifying dependent labels explicitly from the dataset. In this paper we propose new approaches for identifying and modeling existing dependencies between labels. One principal contribution of this work is a theoretical confirmation of the reduction in sample complexity that is gained from unconditional dependence. Additionally, we develop methods for identifying conditionally and unconditionally dependent label pairs; clustering them into several mutually exclusive subsets; and finally, performing multi-label classification incorporating the discovered dependencies. We compare these two notions of label dependence (conditional and unconditional) and evaluate their performance on various benchmark and artificial datasets. We also compare and analyze labels identified as dependent by each of the methods. Moreover, we define an ensemble framework for the new methods and compare it to existing ensemble methods. An empirical comparison of the new approaches to existing base-line and state-of-the-art methods on 12 various benchmark datasets demonstrates that in many cases the proposed single-classifier and ensemble methods outperform many multi-label classification algorithms. Perhaps surprisingly, we discover that the weaker notion of unconditional dependence plays the decisive role.     

H2O2-Diffusion Through Liposomes

The toxicity of H₂O₂ suggests that it readily diffuses through cell membranes. The present paper demonstrates via a chromogenic peroxidase reaction that H₂O₂ can cross the lipid bilayer membrane without affecting the cell wall's integrity. In this demonstration, the chromogen ABTS was encapsulated with horseradish peroxidase into ovolecithin multilamellar liposomes. Upon the addition of dilute H₂O₂, the color was generated exclusively within the intact liposomes. Related results were observed with the AAP/DHBS chromogenic couple; however, this system is complicated by the small holes created by the phenol DHBS in the liposomal wall.     

Lithium?Oxygen Electrochemistry in Non-Aqueous Solutions

Pairing lithium and oxygen in aprotic solvents can theoretically lead to one of the most promising electrochemical cells available. If successful, this system could compete with technologies such as the internal combustion engine and provide an energy density that can accommodate electric vehicle demands. However, there are many problems that have inhibited this technology from becoming realistic. One of the main reasons is capacity fading after only a few cycles, which is caused by the instability of electrolyte solutions in the presence of reduced oxygen species like O₂^.− and O₂²⁻. In recent years, using various analytical tools, researchers have been able to isolate the breakdown products arising from the reactions occurring between the aprotic solvent and the reduced oxygen species. Nevertheless, no solvents have yet been found that are fully stable throughout the reduction and oxidation processes. However, an understanding of these decomposition mechanisms can help us in designing new systems that are more stable toward the aggressive conditions taking place in Li O₂ cell operation. This review will include analytical studies on the most widely used solvents in current Li O₂ research.     

Reaction of Coumarins with Superoxide Anion Radical (O2): Facile Entry to o-Coumarinic Acid Systems

Coumarins 1–5 were reacted with O₂ in aprotic media. The course of the reaction was followed by using methyl iodide to “Cap” (trap) the oxy-anions generated. Abstraction of the enolic hydrogen (where available) proved to be the most facile process followed by “saponification” of the lactone linkage. The cis-stereochemistry of the original coumarin was maintained allowing easy access to a variety of o-coumarinic acid systems.     

A VLSI field-programmable mixed-signal array to perform neural signal processing and neural modeling in a prosthetic system

A very-large-scale integration field-programmable mixed-signal array specialized for neural signal processing and neural modeling has been designed. This has been fabricated as a core on a chip prototype intended for use in an implantable closed-loop prosthetic system aimed at rehabilitation of the learning of a discrete motor response. The chosen experimental context is cerebellar classical conditioning of the eye-blink response. The programmable system is based on the intimate mixing of switched capacitor analog techniques with low speed digital computation; power saving innovations within this framework are presented. The utility of the system is demonstrated by the implementation of a motor classical conditioning model applied to eye-blink conditioning in real time with associated neural signal processing. Paired conditioned and unconditioned stimuli were repeatedly presented to an anesthetized rat and recordings were taken simultaneously from two precerebellar nuclei. These paired stimuli were detected in real time from this multichannel data. This resulted in the acquisition of a trigger for a well-timed conditioned eye-blink response, and repetition of unpaired trials constructed from the same data led to the extinction of the conditioned response trigger, compatible with natural cerebellar learning in awake animals.     

The effects of electromagnetic coupling on MMIC design

The undesired effects of electromagnetic coupling between passive elements in MIC and MMIC designs are examined. Measured data of several strongly coupled circuit topologies commonly encountered in MMIC design are compared to simulations using circuit theory programs employing discontinuity models as well as electromagnetic field simulation tools. The results of this comparison are used to derive a set of specifications for a simulation tool suitable for use in the design of MIC and MMIC circuits.     

Clinical evaluation of MK-185: A new hypolipidemic drug

A double-blind crossover study in 11 patients suggests that MK-185 may be a useful drug in lowering plasma triglyceride levels, whereas it has minimal effects on cholesterol levels. A good correlation between the hypouricemic and protein bound iodine lowering effects and plasma level of MK-185 was seen. No correlation with the hypolipidemic effect was noted. One of eight papers presented at the symposium “Recent Advances in Drugs Affecting Lipid Metabolism,” AOCS Meeting, Houston, May 1971.     

Apportioned margin approach for cost sensitive large margin classifiers

We consider the problem of cost sensitive multiclass classification, where we would like to increase the sensitivity of an important class at the expense of a less important one. We adopt an apportioned margin framework to address this problem, which enables an efficient margin shift between classes that share the same boundary. The decision boundary between all pairs of classes divides the margin between them in accordance with a given prioritization vector, which yields a tighter error bound for the important classes while also reducing the overall out-of-sample error. In addition to demonstrating an efficient implementation of our framework, we derive generalization bounds, demonstrate Fisher consistency, adapt the framework to Mercer’s kernel and to neural networks, and report promising empirical results on all accounts.

     

Deciding unique decodability of bigram counts via finite automata

We revisit the problem of deciding by means of a finite automaton whether a string is uniquely decodable from its bigram counts. An efficient algorithm for constructing a polynomial-size Nondeterministic Finite Automaton (NFA) that decides unique decodability is given. This NFA may be simulated efficiently in time and space. Conversely, we show that the minimum deterministic finite automaton for deciding unique decodability has exponentially many states in alphabet size, and compute the correct order of magnitude of the exponent.     

Boosting conditional probability estimators

In the standard agnostic multiclass model, <instance, label > pairs are sampled independently from some underlying distribution. This distribution induces a conditional probability over the labels given an instance, and our goal in this paper is to learn this conditional distribution. Since even unconditional densities are quite challenging to learn, we give our learner access to <instance, conditional distribution > pairs. Assuming a base learner oracle in this model, we might seek a boosting algorithm for constructing a strong learner. Unfortunately, without further assumptions, this is provably impossible. However, we give a new boosting algorithm that succeeds in the following sense: given a base learner guaranteed to achieve some average accuracy (i.e., risk), we efficiently construct a learner that achieves the same level of accuracy with arbitrarily high probability. We give generalization guarantees of several different kinds, including distribution-free accuracy and risk bounds. None of our estimates depend on the number of boosting rounds and some of them admit dimension-free formulations.