preconstruct

 1"""
 2The `preconstruct` module creates a dataset that is arranged to facilitate
 3neural decoding, i.e. reconstructing a stimulus given the neural
 4response that it induces.
 5
 6To create the dataset, you must create a `preconstruct.sources.DataSource`
 7and pass it a `preconstruct.dataset.DatasetBuilder` and configure your dataset.
 8
 9## What this library can do
10
11- Download your [pprox](https://meliza.org/spec:2/pprox/) files
12and stimuli in WAVE format from [Neurobank](https://github.com/melizalab/neurobank)
13(`preconstruct.sources.NeurobankSource`)
14- Allow you to easily switch between different sources of data, such as
15local files (`preconstruct.sources.FsSource`)
16or Python dictionaries (`preconstruct.sources.MemorySource`)
17- Create gammatone-based spectrogram from raw WAVE data of the stimuli
18(`preconstruct.dataset.DatasetBuilder.add_stimuli`)
19- Automatically update your pprox files to match the
20[stimtrial specification](https://github.com/melizalab/lab_specs/blob/main/specs/stimulus_trial.md)
21and store the data in a convenient pandas DataFrame
22- Convert the point process from the pprox data into an array of fixed duration
23bins containing the count of spikes occuring within the corresponding time window
24(`preconstruct.dataset.DatasetBuilder.bin_responses`)
25- Project the spikes into an alternate basis (`preconstruct.basisfunctions`)
26- Rearrange the binned pprox data into windows of a given duration `tau` for
27input into a neural decoding model
28"""
29from joblib import Memory
30from appdirs import user_cache_dir
31
32from pathlib import Path
33from single_source import get_version
34
35__version__ = get_version(__name__, Path(__file__).parent.parent)
36
37APP_NAME = "preconstruct"
38APP_AUTHOR = "melizalab"
39
40
41_cache_dir = user_cache_dir(APP_NAME, APP_AUTHOR)
42_mem = Memory(_cache_dir, verbose=0)
43
44from preconstruct.dataset import Dataset, DatasetBuilder
45from preconstruct import basisfunctions