Adaptive decoding for brain-machine interfaces through Bayesian parameter updates

Brain-machine interfaces (BMIs) transform the activity of neurons recorded in motor areas of the brain into movements of external actuators. Representation of movements by neuronal populations varies over time, during both voluntary limb movements and movements controlled through BMIs, due to motor learning, neuronal plasticity, and instability in recordings. To ensure accurate BMI performance over long time spans, BMI decoders must adapt to these changes. We propose the Bayesian regression self-training method for updating the parameters of an unscented Kalman filter decoder. This novel paradigm uses the decoder's output to periodically update its neuronal tuning model in a Bayesian linear regression. We use two previously known statistical formulations of Bayesian linear regression: a joint formulation, which allows fast and exact inference, and a factorized formulation, which allows the addition and temporary omission of neurons from updates but requires approximate variational inference. To evaluate these methods, we performed offline reconstructions and closed-loop experiments with rhesus monkeys implanted cortically with microwire electrodes. Offline reconstructions used data recorded in areas M1, S1, PMd, SMA, and PP of three monkeys while they controlled a cursor using a handheld joystick. The Bayesian regression self-training updates significantly improved the accuracy of offline reconstructions compared to the same decoder without updates. We performed 11 sessions of real-time, closed-loop experiments with a monkey implanted in areas M1 and S1. These sessions spanned 29 days. The monkey controlled the cursor using the decoder with and without updates. The updates maintained control accuracy and did not require information about monkey hand movements, assumptions about desired movements, or knowledge of the intended movement goals as training signals. These results indicate that Bayesian regression self-training can maintain BMI control accuracy over long periods, making clinical neuroprosthetics more viable.     

Research support system for stochastic data processing

The paper describes a research support system named “MSM Tools” that can be used for stochastic modelling of real processes in various information systems and implements the heterogeneous computing paradigm. The proposed approach to data mining is based on method of moving separation of probability mixtures. To obtain statistical estimations of the unknown parameters of mixed probability models, the system implements several modifications of so-called EM algorithm (including grid modifications for the NVIDIA CUDA architecture) that are commonly used in such areas as pattern recognition, clustering, classification, processing of censored and/or truncated data. An example of real data analysis via “MSM Tools” service is given.     

Pattern matching in the Hamming distance with thresholds

It has long been known that pattern matching in the Hamming distance metric can be done in time, where n is the length of the text, m is the length of the pattern, and Σ is the alphabet. The classic algorithm for this is due to Abrahamson and Kosaraju. This paper considers the following generalization, motivated by the situation where the entries in the text and pattern are analog, or distorted by additive noise, or imprecisely given for some other reason: in any alignment of the pattern with the text, two aligned symbols a and b contribute +1 to the similarity score if they differ by no more than a given threshold θ, otherwise they contribute zero. We give an time algorithm for this more general version of the problem; the classic Hamming distance matching problem is the special case of θ=0.     

A Newton-like finite element scheme for compressible gas flows

Semi-implicit and Newton-like finite element methods are developed for the stationary compressible Euler equations. The Galerkin discretization of the inviscid fluxes is potentially oscillatory and unstable. To suppress numerical oscillations, the spatial discretization is performed by a high-resolution finite element scheme based on algebraic flux correction. A multidimensional limiter of TVD type is employed. An important goal is the efficient computation of stationary solutions in a wide range of Mach numbers, which is a challenging task due to oscillatory correction factors associated with TVD-type flux limiters. A semi-implicit scheme is derived by a time-lagged linearization of the nonlinear residual, and a Newton-like method is obtained in the limit of infinite CFL numbers. Special emphasis is laid on the numerical treatment of weakly imposed characteristic boundary conditions. Numerical evidence for unconditional stability is presented. It is shown that the proposed approach offers higher accuracy and better convergence behavior than algorithms in which the boundary conditions are implemented in a strong sense.     

Fluorescence lifetime imaging of coral fluorescent proteins

Corals, like many other coelenterates, contain fluorescent pigments that show considerable homology with the well known green fluorescent protein of the jellyfish Aequoria. In corals, unlike jellyfish, multiple proteins are present and the range of excitations and emissions suggest the possibility of energy transfer. The occurrence of F?rster resonant energy transfer (FRET) between fluorescent proteins in corals has already been reported and time-resolved spectra have shown the effect on fluorescent lifetime, but without any spatial resolution. Lifetime confocal microscopy offers lower time resolution but excellent spatial resolution. Lifetimes of the isolated A. millepora pigments amilFP490, amilFP504, and amilFP593 (names indicate emission peaks) were 2.8, 2.9, and 2.9 ns, respectively. In the coral sample, imaging the entire emission spectrum from 420 nm, the mean lifetime was reduced to 1.5 ns, implying that FRET was occurring. Looking just at the fluorescence from FRET donors the lifetime was even shorter, at 1.3 ns, supporting this interpretation. In contrast, no reduction in lifetime is seen in the coral Euphyllia ancora, where the pigment distribution also suggests that the pigments are unlikely to be involved in photoprotection. This study set out to determine the extent of FRET between pigments in two corals, Acropora millepora and Euphyllia, ancora which differ in the arrangement of their pigments and hence possibly in pigment function.     

Online variance minimization

We consider the following type of online variance minimization problem: In every trial t our algorithms get a covariance matrix C ^t and try to select a parameter vector w ^t−1 such that the total variance over a sequence of trials ?_t=1^T (w^t-1)^T C^tw^t-1\sum_{t=1}^{T} (\boldsymbol {w}^{t-1})^{\top} \boldsymbol {C}^{t}\boldsymbol {w}^{t-1} is not much larger than the total variance of the best parameter vector u chosen in hindsight. Two parameter spaces in ℝ ⁿ are considered—the probability simplex and the unit sphere. The first space is associated with the problem of minimizing risk in stock portfolios and the second space leads to an online calculation of the eigenvector with minimum eigenvalue of the total covariance matrix ?_t=1^T C^t\sum_{t=1}^{T} \boldsymbol {C}^{t}. For the first parameter space we apply the Exponentiated Gradient algorithm which is motivated with a relative entropy regularization. In the second case, the algorithm has to maintain uncertainty information over all unit directions u. For this purpose, directions are represented as dyads uu ^⊤ and the uncertainty over all directions as a mixture of dyads which is a density matrix. The motivating divergence for density matrices is the quantum version of the relative entropy and the resulting algorithm is a special case of the Matrix Exponentiated Gradient algorithm. In each of the two cases we prove bounds on the additional total variance incurred by the online algorithm over the best offline parameter.     

GPR55 Receptor Activation by the N-Acyl Dopamine Family Lipids Induces Apoptosis in Cancer Cells via the Nitric Oxide Synthase (nNOS) Over-Stimulation

GPR55 is a GPCR of the non-CB1/CB2 cannabinoid receptor family, which is activated by lysophosphatidylinositol (LPI) and stimulates the proliferation of cancer cells. Anandamide, a bioactive lipid endocannabinoid, acts as a biased agonist of GPR55 and induces cancer cell death, but is unstable and psychoactive. We hypothesized that other endocannabinoids and structurally similar compounds, which are more hydrolytically stable, could also induce cancer cell death via GPR55 activation. We chemically synthesized and tested a set of fatty acid amides and esters for cell death induction via GPR55 activation. The most active compounds appeared to be N-acyl dopamines, especially N-docosahexaenoyl dopamine (DHA-DA). Using a panel of cancer cell lines and a set of receptor and intracellular signal transduction machinery inhibitors together with cell viability, Ca²⁺, NO, ROS (reactive oxygen species) and gene expression measurement, we showed for the first time that for these compounds, the mechanism of cell death induction differed from that published for anandamide and included neuronal nitric oxide synthase (nNOS) overstimulation with concomitant oxidative stress induction. The combination of DHA-DA with LPI, which normally stimulates cancer proliferation and is increased in cancer setting, had an increased cytotoxicity for the cancer cells indicating a therapeutic potential.     

Scientometric Methods for Identifying Scientific Leaders: New Mathematical Models

A new approach to the multidimensional assessment of research organizations is introduced. Based on the use of quantitative data, various types of research and development results are considered in the approach.