Recent years have seen dramatic increase in applications of deep learning techniques to solve real world problem, range from social network to online ecommerce etc. However, the intelligence system powered by deep learning itself is vulnerable and lucrative target to attackers. In 2014, Christian Szegedy et al. firstly found that the highly accurate modern deep learning models are susceptible to adversarial samples that are derived from the original images by adding small perturbations. To address the security problem of AI system, many methods for model attack and defense have been proposed. However, the problem is far from being solved.
The IJCAI-2019 Alibaba Adversarial AI Challenge (AAAC2019) aims at providing a venue for AI practitioners to explore the security of the AI models. Our competition this year focus on image classification tasks, and includes model attacks and model defenses. The participants can play as an attacker to fool our models, or can play as a defender to provide robust model against the adversarial samples.
Three tasks are proposed in this competition including non-targeted attack, targeted attack, and model defense. Specifically,
Different from previous competitions, it is the first time that utilizes the images from online e-commerce as the underlining dataset. Totally, 110,000 product images, which come from 110 categories, will be released. The participants can use these data to train more robust defense models or generate adversarial samples with higher quality.
We plan to use the data competition platform Tianchi developed by aliCloud (part of the Group). Since 2014, More than 100 competitions have been successfully hosted on Tianchi, which gathered 250,000 players from 93 countries and regions. This platform is well developed, tested and can be tailored to this contest as needed.
February 15 - May 31, 2018
The competition webpage has been released at https://tianchi.aliyun.com/markets/tianchi/ijcai2019
The Automated Negotiating Agent Competition (ANAC) brings together researchers from the negotiation community and provides a unique benchmark for evaluating practical negotiation strategies in multi-issue domains. The ANAC has the following aims:
The previous competitions have spawned novel research in AI in the field of autonomous agent design which are available to the wider research community. This year, we would like to introduce a variety of negotiation research challenges:
We expect innovative and novel agent strategies will be developed, and the submitted ANAC 2019 agents will serve as a negotiating agent repository to the negotiation community. The researchers can develop novel negotiating agents and evaluate their agents by comparing their performance with the performance of the ANAC 2019 agents.
February 20th - May 20th, 2019
For more details, please visit the competition webpage : http://web.tuat.ac.jp/~katfuji/ANAC2019/