Cyrus Omar

Neural Correlation Is Stimulus Modulated by Feedforward Inhibitory Circuitry

Correlated variability of neural spiking activity has important consequences for signal processing. How incoming sensory signals shape correlations of population responses remains unclear. Cross-correlations between spiking of different neurons may be particularly consequential in sparsely firing neural populations such as those found in layer 2/3 of sensory cortex. In rat whisker barrel cortex, we found that pairs of excitatory layer 2/3 neurons exhibit similarly low levels of spike count correlation during both spontaneous and sensory-evoked states. The spontaneous activity of excitatory–inhibitory neuron pairs is positively correlated, while sensory stimuli actively decorrelate joint responses. Computational modeling shows how threshold nonlinearities and local inhibition form the basis of a general decorrelating mechanism. We show that inhibitory population activity maintains low correlations in excitatory populations, especially during periods of sensory-evoked coactivation. The role of feedforward inhibition has been previously described in the context of trial-averaged phenomena. Our findings reveal a novel role for inhibition to shape correlations of neural variability and thereby prevent excessive correlations in the face of feedforward sensory-evoked activation.

A feedback information-theoretic approach to the design of brain-computer interfaces

This article presents a new approach to designing brain–computer interfaces (BCIs) that explicitly accounts for both the uncertainty of neural signals and the important role of sensory feedback. This approach views a BCI as the means by which users communicate intent to an external device and models intent as a string in an ordered symbolic language. This abstraction allows the problem of designing a BCI to be reformulated as the problem of designing a reliable communication protocol using tools from feedback information theory. Here, this protocol is given by a posterior matching scheme. This scheme is not only provably optimal but also easily understood and implemented by a human user. Experimental validation is provided by an interface for text entry and an interface for tracing smooth planar curves, where input is taken in each case from an electroencephalograph during left- and right-hand motor imagery.

Safely Composable Type-Specific Languages

Programming languages often include specialized syntax for common datatypes e.g. lists and some also build in support for specific specialized datatypes e.g. regular expressions, but user-defined types must use general-purpose syntax. Frustration with this causes developers to use strings, rather than structured data, with alarming frequency, leading to correctness, performance, security, and usability issues. Allowing library providers to modularly extend a language with new syntax could help address these issues. Unfortunately, prior mechanisms either limit expressiveness or are not safely composable: individually unambiguous extensions can still cause ambiguities when used together. We introduce type-specific languages TSLs: logic associated with a type that determines how the bodies of generic literals, able to contain arbitrary syntax, are parsed and elaborated, hygienically. The TSL for a type is invoked only when a literal appears where a term of that type is expected, guaranteeing non-interference. We give evidence supporting the applicability of this approach and formally specify it with a bidirectionally typed elaboration semantics for the Wyvern programming language.

Collaborative infrastructure for test-driven scientific model validation

One of the pillars of the modern scientific method is model validation: comparing a scientific models predictions against empirical observations. Today, a scientist demonstrates the validity of a model by making an argument in a paper and submitting it for peer review, a process comparable to code review in software engineering. While human review helps to ensure that contributions meet high-level goals, software engineers typically supplement it with unit testing to get a more complete picture of the status of a project. We argue that a similar test-driven methodology would be valuable to scientific communities as they seek to validate increasingly complex models against growing repositories of empirical data. Scientific communities differ from software communities in several key ways, however. In this paper, we introduce SciUnit, a framework for test-driven scientific model validation, and outline how, supported by new and existing collaborative infrastructure, it could integrate into the modern scientific process.

Querying the user properly for high-performance brain-machine interfaces: Recursive estimation, control, and feedback information-theoretic perspectives

We propose a complementary approach to the design of neural prosthetic interfaces that goes beyond the standard approach of estimating desired control signals from neural activity. We exploit the fact that the for a users intended application, the dynamics of the prosthetic in fact impact subsequent desired control inputs. We illustrate that changing the dynamic response of a prosthetic device can make specific tasks significantly easier to accomplish. Our approach relies upon principles from stochastic control and feedback information theory, and we illustrate its effectiveness both theoretically and experimentally - in terms of spelling words from a menu of characters using binary surface electromyography classification.

Structured statistical syntax tree prediction

Statistical models of source code can be used to improve code completion systems, assistive interfaces, and code compression engines. We are developing a statistical model where programs are represented as syntax trees, rather than simply a stream of tokens. Our model, initially for the Java language, combines corpus data with information about syn- tax, types and the program context. We tested this model using open source code corpuses and find that our model is significantly more accurate than the current state of the art, providing initial evidence for our claim that combining structural and statistical information is a fruitful strategy.

Type-directed, whitespace-delimited parsing for embedded DSLs

Domain-specific languages improve ease-of-use, expressiveness and verifiability, but defining and using different DSLs within a single application remains difficult. We introduce an approach for embedded DSLs where 1) whitespace delimits DSL-governed blocks, and 2) the parsing and type checking phases occur in tandem so that the expected type of the block determines which domain-specific parser governs that block. We argue that this approach occupies a sweet spot, providing high expressiveness and ease-of-use while maintaining safe composability. We introduce the design, provide examples and describe an ongoing implementation of this strategy in the Wyvern programming language. We also discuss how a more conventional keyword-directed strategy for parsing of DSLs can arise as a special case of this type-directed strategy.

Policies for neural prosthetic control: Initial experiments with a text interface

Neural interfaces use estimates of brain or muscle activity to generate control inputs for a prosthetic device. Most previous work focuses on estimating neural activity more accurately. This paper focuses on generating bette control inputs. It shows that changing the dynamic response of a prosthetic device can make specific tasks easier to accomplish. It also presents experimental results for which neural activity is measured using surface electromyography, the prosthetic is a 1-D cursor, and the task is to spell words from a menu of characters.

Programmable semantic fragments: the design and implementation of typy

This paper introduces typy, a statically typed programming language embedded by reflection into Python. typy features a fragmentary semantics, i.e. it delegates semantic control over each term, drawn from Pythons fixed concrete and abstract syntax, to some contextually relevant user-defined semantic fragment. The delegated fragment programmatically 1) typechecks the term (following a bidirectional protocol); and 2) assigns dynamic meaning to the term by computing a translation to Python. We argue that this design is expressive with examples of fragments that express the static and dynamic semantics of 1) functional records; 2) labeled sums (with nested pattern matching a la ML); 3) a variation on JavaScripts prototypal object system; and 4) typed foreign interfaces to Python and OpenCL. These semantic structures are, or would need to be, defined primitively in conventionally structured languages. We further argue that this design is compositionally well-behaved. It avoids the expression problem and the problems of grammar composition because the syntax is fixed. Moreover, programs are semantically stable under fragment composition (i.e. defining a new fragment will not change the meaning of existing program components.)

Composable and hygienic typed syntax macros

Syntax extension mechanisms are powerful, but reasoning about syntax extensions can be difficult. Recent work on type-specific languages (TSLs) addressed reasoning about composition, hygiene and typing for extensions introducing new literal forms. We supplement TSLs with typed syntax macros (TSMs), which, unlike TSLs, are explicitly invoked to give meaning to delimited segments of arbitrary syntax. To maintain a typing discipline, we describe two avors of term-level TSMs: synthetic TSMs specify the type of term that they generate, while analytic TSMs can generate terms of arbitrary type, but can only be used in positions where the type is otherwise known. At the level of types, we describe a third avor of TSM that generates a type of a specified kind along with its TSL and show interesting use cases where the two mechanisms operate in concert.