Kaggle Community

[ankitjha97]

India has the 2nd highest community in Kaggle and it is a great and dynamic platform to learn.
Those who want to participate in kaggle competitions and kernels can reply to this issue so that we all work together as a team because it will be very easy to learn and get popular on kaggle if we participate together. In this way, our college will gain a good reputation in Data Science and Machine Learning which can attract good companies.

[ayush1999]

Sounds great. Also, it'd be better if we could arrange them in the labs for a few hours. LAN and GPUs(if a few computers in the lab have it) could be used for learning and participating.

[shivank01]

Yeah, It will be great. I want to be a part of this.

[ankitjha97]

@ayush1999 , @shivank01 , could you folks please post it on slack so that we can have a dedicated group to it in the telegram. I am not there on slack.

[shivank01]

@ankitjha97 I have posted

[saiamrit]

Good initiative @ankitjha97 ..I am in too.

[saiamrit]

If its not possible to arrange systems then we can atleast ask for funds and buy some credits on cloud services.there are many like one is PaperSpace which provides completely configured linux virtual machines with GPU.You all can go through the pricing and details here ( https://www.paperspace.com/ml )

Paperspace

Cloud Machine Learning, AI, and effortless GPU infrastructure

[shivank01]

@saiamrit There is one more floydhub. It provides few hours of free GPU instance as well.

[saiamrit]

Ya I know about floydhub and have seen the free package but the storage in the free package is less like 10gb. Its fine for small data at personal level but when we are planning for the college, a higher storage is required and floydhub is monthly billing but PaperSpace has monthly as well as hourly plan too.Any ways we can have these discussions on the telegram group.Lets form the group and dive into these discussions.

[saiamrit]

Guys someone please create a telegram group with people who replied to the slack thread.Lets have some discussion within one two days about the infrastructure requirements and other plans to encourage the budding Kagglers.

[arijeetsat]

@saiamrit , @hdidwania Can you take this discussion ahead, and put some update for it here.

[hdidwania]

Can you take this discussion ahead, and put some update for it here.

Cool. Lets see what we can do for this.
We certainly need to act as a team (for the competitions are huge, tough and generally research based). Getting access to machines/cloud is a must for participating. We as an ML community can publish 2-3 kernels weekly which others can explore, and even carry on some workshop-ish like event where we can together work on a kernel, learning something useful.

[arijeetsat]

Yeah great

[piyushmishra12]

I want to be a part of it too!

[sudoankit]

Just a reminder, IIIT BBS has invested heavy on Microsoft Azure Cloud, ask for Azure ML if possible to the authorities. I doubt we can train students on Kaggle without dedicated hardware, minimum a GTX 1070 for vision tasks and >= GTX1060 for NLP, etc.

[dnabanita7]

@saiamrit Can I get updates on this? What about using spotty for training DL models which deploys it on AWS? It costs only Rs.2/- for registration and creating a bucket in AWS. Each spot instance costs less than 2 or 3 rupees. Here is the user guideline.