bnpm 0.1.0

A library of useful modules for data analysis.

basic_neural_processing_modules

Personal library of functions used in analyzing neural data. If you find a bug or just want to reach out: [e-mail]

Installation

A note on requirements: There are a lot of different libraries that are used in this package. As such there is a basic requirements.txt file that will install all the absolutely required libraries. However, there are other libraries that are imported in specific functions. These are listed in requirements_advanced.txt. I recommend installing only the libraries that you need for your specific project, as some require specific versions of other libraries or operating systems or GPUs.

1. Clone the repository

   • cd path/to/your/preferred/directory
   • git clone https://github.com/RichieHakim/basic_neural_processing_modules

2. Install requirements:

   • cd basic_neural_processing_modules
   • pip install -e .
   • Install any other libraries in requirements_advanced.txt that you need for your project.

3. Install the package:

   • cd path/to/basic_neural_processing_modules
   • pip install -e .

you can now import with: import bnpm

Usage

My favorites:

• featurization.Toeplitz_convolution2d
  • 1D and 2D convolution. Uses sparse Toeplitz matrix multiplication to speed up computation.
  • Allows for sparse convolution.
  • Same options as scipy.signal.convolve2d, but orders of magnitude faster in most cases.
• spectral.VQT
  • Variable-Q transform. Generates spectrograms with variable frequency resolution.
  • Comparable to librosa's VQT, but faster, more flexible, without approximations, with GPU support, and pytorch autograd compatible.

Other useful functions:

• Signal Processing:

  • timeSeries.rolling_percentile_rq_multicore
    • Fast rolling percentile calculation
  • timeSeries.event_triggered_traces
    • Fast creation of a matrix of aligned traces relative to specified event times

• Machine Learning:

  • decomposition.torch_PCA
    • Fast standard PCA using PyTorch
  • linear_regression.LinearRegression_sweep
    • Performs linear regression with a variety of hyperparameters and methods (L1, L2, Logistic, optional GPU methods using cuml)
  • misc.make_batches
    • Creates batches of data or any other iterable
  • similarity.orthogonalize and similarity.pairwise_orthogonalization
    • Orthogonalize a matrix relative to a set of vectors using a Gram-Schmidt related process

• Miscellaneous

  • path_helpers.find_paths
    • Find paths to files and/or folders in a directory. Searches recursively using regex.
  • misc.estimate_size_of_float_array
    • Estimates the size of a float array in bytes
  • image_processing.play_video_cv2
    • Plays and/or saves a 3D array as a video using OpenCV
  • h5_handling.simple_save and h5_handling.simple_load
    • Simple lazy loading and saving of dictionaries as nested h5 files
  • parallel_helpers.multiprocessing_pool_along_axis
    • Easy parallelization of a function along an axis
  • plotting_helpers.get_subplot_indices
    • Returns the subscript indices of the subplots in a figure