schung039/neural_manifolds_replicamft

neural_manifolds_replicamft's Issues

Install issues: python version and pytorch

I've spent some time recently installing this and found a couple issues and fixes. My typical python version (3.9.12) was yielding a lot of errors while installing the dependencies in a new environment via anaconda. The older numpy and scipy version required conflict with recent python, and I also needed to manually install pytorch and torchvision since I didn't have them and they weren't in the requirements.txt file. So here's my fix for a fresh install that currently works, using packages and python released around July 2019:

My sysem and python versions-
OS - Windows 11
CPU - Intel Core i9-12900K, 3200 Mhz, 16 Core(s), 24 Logical Processor(s)
GPU - NVIDIA 3080
pyhon version - 3.9.12
conda version - 4.14.0

My current install method

I downloaded and extracted the code repo ("neural_manifolds_replicaMFT-master").

In the conda prompt (anaconda.org), navigate to the location of the "\neural_manifolds_replicaMFT-master" folder containing the requirements.txt file

Create a new environment with python 3.6.9

(base) conda create -n py369 python=3.6.9 anaconda
(base) conda activate py369

(py369) pip install -r requirements.txt
(py369) pip install -e .

Install previous pytorch and torchvision releases, pytorch==1.1.0 torchvision==0.3.0 didn't work

(py369) conda install pytorch==1.5.1 torchvision==0.6.1 -c pytorch

Open up jupyter and the examples should work

(py369) jupyter notebook

Analysis fails for small P

Hello, thank you for this great work!

In my hand the analysis seems to fail for small P.
reusable example:

np.random.seed(0)
X = [np.random.randn(5000, 50) for i in range(3)]
kappa = 0
n_t = 200
capacity_all, radius_all, dimension_all, center_correlation, K = manifold_analysis_corr(X, kappa, n_t)

Which leads to an assertion error in fun_FA

This seems to come from the setting of maxK (line 62 of manifold_analysis_correlation):
maxK = np.argmax([t if t < 0.95 else 0 for t in total]) +11
as changing it to
maxK = np.argmax([t if t < 0.95 else 0 for t in total]) +1
solves the issue for P>2

The issue remains for P=2.

Overall is the technique adaptable for comparison of 2 manifolds, and similarly, does it make sense to try using it for low number of class (<100)?

Thank you for your help!

juptyer kernel died when I run manifold_analysis_corr

I configured the virtual environment according to this issue #8.

pywinpty supports at least 3.7, so it is not just a plug-in problem that prevents the installation of jupyter, and I upgraded to 3.7.12

Then I run

import numpy as np
from mftma.manifold_analysis_correlation import manifold_analysis_corr

np.random.seed(0)
X = [np.random.randn(5000, 50) for i in range(100)] 
kappa = 0
n_t = 200

capacity_all, radius_all, dimension_all, center_correlation, K = manifold_analysis_corr(X, kappa, n_t)

my jupter kernel will die (VScode and jupter notebook), I don't know what's going on and I'd like to ask for your help