Research Scientist @Nokia @BellLabs. AI/ML for data-efficient mobile sensing. PhD candidate at @Cambridge_Uni
Joined July 2019
- Tweets8
- Following23
- Followers33
- Likes12
1/5 Tired of self-supervised models that can't adapt to new data?
In our #WACV2024 paper, we propose Kaizen - an end-to-end approach for continual learning that performs knowledge distillation across both the pre-training and fine-tuning steps.
👉 arxiv.org/abs/2303.17235
4/5 We evaluate on class-incremental CIFAR-100 and ImageNet-100 across multiple metrics, and Kaizen outperforms previous state-of-the-art CSSL methods. Analyzing per-task performance shows that Kaizen strikes a balance between retaining old knowledge and learning new tasks.
5/5 In longer continual learning scenarios, Kaizen outperforms in the first tasks, but matches baselines later as it prioritizes retention. Stay tuned for code release 🔜!
Shout out to the great team @LorenaQendro @spdimitris @raswak @cecim @akhilmathurs! See you at #WACV2024 🏄
I am co-organising SOAR, a tutorial on SOlving the sensor-based Activity Recognition problem at UbiComp tomorrow! Come join us if you're attending UbiComp/ISWC 2023. #UbiComp2023 #ISWC2023 - more details at sites.google.com/view/soar-t…
Ian Tang retweeted
Super excited to announce the Human-Centric Representation Learning workshop at #AAAI2024!
Our workshop will bring together people working on generalizable models that take into account the unique properties of human-centric data.
📅 Deadline: 24/11
🔗 hcrl-workshop.github.io/2024…
Ian Tang retweeted
We had a wonderful showcase day for the renewal of our Centre today! thanks @BellLabs for the support. @Cambridge_CL @arberesford @raswak @thierry_klein
