HN user

mostafab

141 karma
Posts24
Comments39
View on HN
medium.com 7y ago

You did not need deep learning to generate molecules

mostafab
3pts0
medium.com 8y ago

Response to Harvard Professor Alan Aspuru-Guzik

mostafab
1pts1
medium.com 8y ago

Reply to Vijay Pande, VC Investor and Stanford Professor

mostafab
1pts1
medium.com 8y ago

Response to Harvard Professor Alan Aspuru-Guzik

mostafab
1pts0
medium.com 8y ago

Reply to Vijay Pande, VC Investor and Stanford Professor

mostafab
1pts0
blogs.sciencemag.org 8y ago

Objections to (Some) Drug Discovery AI

mostafab
1pts0
medium.com 8y ago

AI in drug discovery is overhyped: examples from AstraZeneca, Harvard, Stanford

mostafab
229pts99
medium.com 8y ago

AI in drug discovery is overhyped: examples from AstraZeneca,Harvard,Stanford

mostafab
2pts0
medium.com 8y ago

AI in drug discovery is overhyped: examples from AstraZeneca, Harvard, Stanford

mostafab
1pts0
medium.com 8y ago

Startcrowd: A Blockchain Network for Artificial Intelligence Projects

mostafab
4pts0
medium.com 8y ago

ChemGAN challenge for drug discovery: can AI reproduce natural diversity?

mostafab
1pts0
medium.com 8y ago

ChemGAN challenge for drug discovery: can AI reproduce natural diversity?

mostafab
1pts0
hackernoon.com 8y ago

Thank You for Inviting Me to Build Your App in Exchange for Equity,Kind Stranger

mostafab
6pts1
www.linkedin.com 9y ago

Remote Jobs in Artificial Intelligence and Data Science

mostafab
3pts0
www.linkedin.com 9y ago

Remote Jobs in Artificial Intelligence and Data Science

mostafab
5pts0
www.facebook.com 9y ago

Remote Jobs in Artificial Intelligence and Data Science

mostafab
3pts0
hackernoon.com 9y ago

Make Pharma Great Again with Artificial Intelligence: Some Challenges

mostafab
1pts0
medium.com 9y ago

Make Pharma Great Again with Artificial Intelligence: Some Challenges

mostafab
1pts0
startcrowd.club 9y ago

Show HN: Startcrowd – Collaborate on ambitious AI products

mostafab
2pts0
hackernoon.com 9y ago

How AI Startups Must Compete with Google: Reply to Fei-Fei Li

mostafab
2pts0
hackernoon.com 9y ago

How AI Startups Must Compete with Google: Reply to Fei-Fei Li

mostafab
3pts0
medium.com 9y ago

How AI Startups Must Compete with Google: Reply to Fei-Fei Li

mostafab
1pts0
medium.com 9y ago

No, Kaggle Is Unsuitable to Study AI and ML. A Reply to Ben Hamner

mostafab
10pts1
medium.com 9y ago

How to start a deep learning startup, Not from scratch: a tutorial

mostafab
1pts0
[dead] 5 months ago

Are people here building Frontier AI Models by Distillation? Did any of you try OpenDistil? Is it better than in-house distillation infrastructure?

This professor gave an argument in his paper. I gave a refutation. If you agree with him that char-CNN was a sophisticated model in early 2017, then you are not well-informed about the situation in deep learning.

The MoleculeNet co-author gave another argument in the comments. My refutation is that you can't claim to lack time after 8*20 man-months have passed.

I am not sure to fully understand your remark, but if you have a benchmark graph convolutions vs. char-CNN, it would be great to write your result and post it on Arxiv. Pande will be interested ;) The problem is not with the theoretical motivation, but with the empirical confirmation.

I also agree that there are many ways to work through the perfect discriminator problem for ORGAN. But it remains to be done (afaik).

From a user viewpoint, Deepchem would greatly benefit from being a better team player with lower-level (Tensorflow) or other (Pytorch) frameworks. The pace of research (in NLP in particular) is too fast to make it realistic to port everything in Deepchem without an unreasonable delay.

Does it fit the Deepchem agenda? That's another question ;)

In general, you are right. But in this particular case, char-CNN is the standard used by many people, Stanford included. It is not sophisticated at all.

Look at the big list of exotic models they tried: http://moleculenet.ai/models

For me, this omission is a negligence. It is selective laziness.

The author (me) does not have 20 PhD students, postdocs and startuppers under his hand, to do the job for Stanford. With my limited resources, it is more cost-effective to do my job against Stanford ;)

Author here

1. I agree that graph convolutions make a lot of sense. However, it is not shown yet that they are better than SMILES, although it might just be a matter of time. A lot of graph variants are possible, or maybe we should look at molecules not as graphs, but as quantum objects (See the other Stanford paper on atom convolutions).

2. Yes, Deepchem is built on top of Tensorflow. This additional layer looks useful, but is it really the case? Using standard NLP models is simpler than using deepchem models, and NLP models might still be the state-of-the-art for chemistry tasks, via SMILES (until we get a reasonably comprehensive benchmark, which MoleculeNet is not).

I am not buying into a chemistry-specific library, until it is shown to be really necessary.