SAiDL
@SforAiDL
Followers
605
Following
146
Media
80
Statuses
356
Society for Artificial Intelligence and Deep Learning. Accelerating student-driven AI research @ BITS Pilani. AI Symposium 23 info : https://t.co/i8FA7jHVjV
Joined June 2020
We are Society for Artificial Intelligence and Deep Learning at @BITSPilaniGoa. Founded a few years ago, we've since tried to create a supportive environment for those interested in research in AI, especially on our campus. (1/6).
1
4
40
These models are called Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior.
1
0
1
The paper reintroduces behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding and domain adaptation.
1
0
1
One reason for the underperformance could be a lack of "behavior tokens'' in LLMs' training corpora as they are removed as noise during data preprocessing. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc.
1
0
1
LLMs, with their wide generalizability, make some progress towards the second level. The third problem deals with predicting and optimizing communication for desired receiver behavior and LLMs are unable to solve this.
1
0
1
Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectiveness. Thanks to telecommunications, the first level problem has produced great advances like the internet.
1
0
0
Check out the Spotlight paper by SAiDL members @ashmitkx, @aditya30502, @someshsingh22, Aanisha Bhattacharyya, Yaman K Singla, Uttaran Bhattacharya, Ishita Dasgupta, Stefano Petrangeli, Rajiv Ratn Shah, Changyou Chen, Balaji Krishnamurthy at ICLR 2024!.
arxiv.org
Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with...
2
2
18
The results also provide strong evidence that making DL models activity sparse and porting them to neuromorphic devices can be a viable strategy that does not compromise on performance.
0
0
1
They achieve up to 20× reduction of computation while maintaining perplexities below 60 on the Penn Treebank language modeling task. Such performance has not been achieved previously with any sparsely activated recurrent neural networks or spiking neural networks
1
0
0
In this work, the authors demonstrate that activity sparsity can compose multiplicatively with parameter sparsity in a recurrent neural network model based on the GRU that is designed to be activity sparse.
1
0
0
Yet, sparse activations, while omnipresent in both biological neural networks and deep learning systems, have not been fully utilized as a compression technique in deep learning. Moreover, the interaction between sparse activations and weight pruning is not fully understood.
1
0
0
Artificial neural networks open up unprecedented ML capabilities at the cost of ever growing computational requirements. Sparsifying the parameters, often achieved through weight pruning, is a powerful technique to compress the model and reduce the computational complexity.
1
0
0
Check out the paper published by SAiDL member.@mukherji_rishav , Mark Schöne, Khaleel Khan, Christian Mayr and @anandsubramoney at the NeurIPS 2023 Workshop on ML with New Compute Paradigms!.
arxiv.org
Artificial neural networks open up unprecedented machine learning capabilities at the cost of ever growing computational requirements. Sparsifying the parameters, often achieved through weight...
1
0
6
Check out the upcoming AAAI 2024 publication by SAiDL member @shreyasbhat23 , Prof.Ashwin Srinivasan, Dr. Tirtharaj Dash, Sowmya Krishnan, Dr. Lovekesh Vig, Dr. Arijit Roy and Prof.Raviprasad Aduri !!.
Thrilled to share our paper "Generating Novel Leads for Drug Discovery using LLMs with Logical Feedback" has been accepted at #AAAI2024 !!🎉. @TCSResearch @appcair . (1/N).
0
0
10
Our session with Varuni Sarwal's (@VaruniSarwal ) and Achint Kumar's (@achint1994 ) talk on Getting started in AI for Science will be starting in 30 minutes at 7:00 PM IST!. WebEx : Join the discussion on our Discord channel:
0
0
0
Our second session of the day, with Nisha Chetana Sastry on Computational Neuroscience, will start in 30 minutes at 3:30 pm IST!. WebEx : Join the discussion on our Discord channel:
0
0
0
Anand Subramoney(@anandsubramoney )'s lecture on "Bio-inspired ML and Scalability" will be starting in 30 minutes at 3:30 PM IST!. WebEx: Join the discussion on our Discord channel:
0
0
1
Our second session of the day, with Moksh Jain(@JainMoksh) on GFlowNets for scientific discovery, will start in 30 minutes at 11:00 am IST!. WebEx: Join the discussion on our Discord channel:
0
0
3
We are excited to announce two of our speakers for "AI Symposium 2023", Varuni Sarwal (@VaruniSarwal), Computer Science PhD at @UCLA, and Achint Kumar(@achint1994), Deep Learning Researcher at @Shell. They will be talking about getting started in AI for Science.
0
0
1