Topographica A general-purpose neural simulator focusing on topographic maps
Download

Topographica Ranking & Summary

Topographica Tags

simulate neural topographic map topographic neural simulator neural map

Topographica Description

A general-purpose neural simulator focusing on topographic maps Topographica is a software package for computational modeling of neural maps, developed by the Neural Networks Research Group at the University of Texas at Austin and the Institute for Adaptive and Neural Computation at the University of Edinburgh.The project is funded by the NIMH Human Brain Project under grant 1R01-MH66991. The goal is to help researchers understand brain function at the level of the topographic maps that make up sensory and motor systems.Topographica is intended to complement the many good low-level neuron simulators that are available, such as Genesis and Neuron. Those simulators focus on modeling the detailed internal behavior of neurons and small networks of them.Topographica instead focuses on the large-scale structure and function that is visible only when many thousands of such neurons are connected into topographic maps containing millions of connections. Many important phenomena cannot be studied without such large networks, including the two-dimensional organization of visual orientation and motion direction maps, and object segmentation and grouping processes.To make such models practical, in Topographica the fundamental unit is a sheet of neurons, rather than a neuron or a part of a neuron. For most simulations, the sheets can be implemented at a high level, consisting of abstract firing-rate or integrate-and-fire neurons. When required for validation or for specific phenomena, Topographica can easily be extended using a Sheet that interfaces to more detailed neuron models in other simulators. Less-detailed sheets can also be used temporarily, e.g. when interacting with the model in real time. Topographica makes it simple to use an appropriate level of complexity and detail, as determined by the available computing power, amount of biological data available for validation and phenomena of interest. What's New in This Release: General improvements: · significant performance improvements in simulations (nearly 2X overall), plotting (2X), and startup time · minor bugfixes · more options for adding noise to ConnectionField shapes · optional line-by-line profiling · optional Cython package for writing fast components Command-line and batch: · -v and -d options to print verbose and debugging messages · new options to run_batch and better progress messages · replaced most commands with ParameterizedFunctions, which have documented, type and bound-checked arguments and allow inheritance of shared functionality · replaced map measurement commands in topo.command with simpler, general-purpose, easily .ty-file controllable versions (see lissom_oo_or.ty and lissom.ty for examples) · save_plotgroup: more useful default values; results can be cached to avoid recomputation · measure_sine_pref: general purpose measurement for any preference that can be tested with a sine grating · Changed locals to script-level parameters using GlobalParams; see examples/lissom.ty · Made gradientplot and fftplot available in batch mode. GUI: · model editor supports non-Sheet EventProcessors and non-CFProjection EPConnections · right-click option for plotting tuning curves · plot windows can be arranged in 2D, not just a row (see Sheet.row_precedence) Example scripts: · example file for interfacing to external simulators (examples/perrinet_retina.ty) · removed outdated or in-progress examples · greatly simplified remaining example scripts · now use GlobalParams to support consistent option setting using -p · allowed saving of local functions and instance method calls in snapshots Component library: · PatternGenerators: Translator; mask_shape parameter also now makes it easy to specify a mask for any pattern, e.g. in the GUI · TransferFns (formerly called OutputFns): HalfRectifyAndPower, Hysteresis, and HomeostaticResponse · Sheets: ActivityCopy · LearningFns: CFPLF_BCMFixed_opt, CFPLF_Scaled_opt · Added HookList parameters to FeatureResponses and LISSOM to make it easier to add user-defined functionality. · Changed names and definitions of various similar concepts (OutputFn, before_presentation, update_command, plot_command, etc.) to reflect shared concept of Hooks and HookLists (lists of callables to run at specific spots in the code). · Parameters: bounds can now be exclusive, optional support for None in most types

Topographica Related Software