search

xiaoyeye / GCNG

using graph convolutional neural network and spaital transcriptomics data to infer cell-cell interactions
56 stars 16 forks source link

Building Model #3

Open WanxiangHuang opened 3 years ago

WanxiangHuang commented 3 years ago

Hi, there are some updates to the Spektral's Convolutional layers:
A_in = Input(tensor=sp_matrix_to_sp_tensor(fltr)) will lead to failure And GraphConv has been removed as well. I will be really appreciated if you can update your code. I am trying to write by myself, but met the following error: Model: "functional_1"

Layer (type) Output Shape Param # Connected to

input_1 (InputLayer) [(None, 3890, 2)] 0

input_2 (InputLayer) [(None, 3890)] 0

gcs_conv (GCSConv) (None, 3890, 32) 128 input_1[0][0] input_2[0][0]

gcs_conv_1 (GCSConv) (None, 3890, 32) 2048 gcs_conv[0][0] input_2[0][0]

flatten (Flatten) (None, 124480) 0 gcs_conv_1[0][0]

dense (Dense) (None, 512) 63734272 flatten[0][0]

dense_1 (Dense) (None, 1) 513 dense[0][0]

Total params: 63,736,961 Trainable params: 63,736,961 Non-trainable params: 0

AssertionError: Could not compute output Tensor("dense_1/Sigmoid:0", shape=(None, 1), dtype=float32)

xiaoyeye commented 3 years ago

Hi, Thanks for your attention and patience. It is really a problem when you using the new Spektral's layers. I suggest use the old one firstly as mentioned in readme. Actually I am pretty busy now, however I and my colleagues are planing to develop new implement ways, and will update the code ASAP.

WanxiangHuang commented 3 years ago

hi, Do you remember the version of Spektral you used before? I used Spektral 0.5.1 , but it still not work. Thanks

xiaoyeye commented 3 years ago

Our cluster has been down for more than 1 month ..... But I will reply you ASAP once it is fixed.

Can you like to try the following: The old version of Spektral, which is based on TensorFlow 1 and the stand-alone Keras library, is still available on the tf1 branch on GitHub and can be installed from source: git clone https://github.com/danielegrattarola/spektral.git cd spektral git checkout develop git checkout tf1 python setup.py install # Or 'pip install .'

QSong-github commented 3 years ago

Our cluster has been down for more than 1 month ..... But I will reply you ASAP once it is fixed.

Can you like to try the following: The old version of Spektral, which is based on TensorFlow 1 and the stand-alone Keras library, is still available on the tf1 branch on GitHub and can be installed from source: git clone https://github.com/danielegrattarola/spektral.git cd spektral git checkout develop git checkout tf1 python setup.py install # Or 'pip install .'

I tried your provided Spektral, however, it shows the following errors:

**from spektral.layers import GraphConv, Dropout Traceback (most recent call last): File "", line 1, in ImportError: cannot import name 'GraphConv' from 'spektral.layers'

Could you kindly help with this?

xiaoyeye commented 3 years ago

Our cluster has been down for more than 1 month ..... But I will reply you ASAP once it is fixed. Can you like to try the following: The old version of Spektral, which is based on TensorFlow 1 and the stand-alone Keras library, is still available on the tf1 branch on GitHub and can be installed from source: git clone https://github.com/danielegrattarola/spektral.git cd spektral git checkout develop git checkout tf1 python setup.py install # Or 'pip install .'

I tried your provided Spektral, however, it shows the following errors:

from spektral.layers import GraphConv, Dropout Traceback (most recent call last): File "", line 1, in ImportError: cannot import name 'GraphConv' from 'spektral.layers' (/Users/qsong/Dropbox/Spatial_interaction/codes/new/lib/python3.7/site-packages/spektral/layers/init.py)

Could you kindly help with this?

Hi， Thanks for your interest. Actually we are buliding a new version based on pytorch, and we release it ASAP and keep you upated. Best

jasperhyp commented 3 years ago

Hello,

Is there an estimate of time that you would release the pytorch version? I have been using torch for a while and transferring to tensorflow might take time, so am expecting that!

Thanks

jasperhyp commented 3 years ago

Also, it looks like this link in your paper has broken down: https://static-content.springer.com/esm/art%3A10.1038%2Fncomms8866/MediaObjects/41467_2015_BFncomms8866_MOESM611_ESM.xlsx ("used as true interactions a curated list of interacting ligands and receptors").

xiaoyeye commented 3 years ago

Hi, Thanks for your interest. I should be able to upload it by the end of April.

xiaoyeye commented 3 years ago

Also, it looks like this link in your paper has broken down: https://static-content.springer.com/esm/art%3A10.1038%2Fncomms8866/MediaObjects/41467_2015_BFncomms8866_MOESM611_ESM.xlsx ("used as true interactions a curated list of interacting ligands and receptors").

You can find the dataset in the following link. https://github.com/xiaoyeye/GCNG/tree/master/LR_database

jasperhyp commented 3 years ago

Thanks!

jasperhyp commented 3 years ago

Hi, Thanks for your interest. I should be able to upload it by the end of April.

Hi, Is there any news about the new version?

xiaoyeye commented 3 years ago

Hi, Thanks for your interest. I should be able to upload it by the end of April.

Hi, Is there any news about the new version?

Hi, I was extremely busy in the last several months. Now I have asked a student to update the new version, and it will come soon. Thanks. Best