Machine Learning Research Staff Member in MIT IBM Watson AI LabCambridge, MA, USA
- Multimodal representation learning, Co-embedding, Joint embedding
- Wasserstein embedding, optimal transport, point clouds, word2cloud
- Logic embedding, hierarchy embedding, combining learning and reasoning, word2ball
- Embedding: structured embedding, metric learning
- Inference-free structured output prediction
- Association learning, link prediction
- Matrix and tensor completion, multi-label classification, zero-shot learning
- Deep structures: Siamese networks, bilinear models, energy based models
- Convex modeling of machine learning problems
- Non-convex global optimization
- Data Dependent Loss Functions
- Generative modeling
- Copula, Quantile functions, Ranking, etc.
Selected Publications and Theses
- Mikhail Yurochkin, Sebastian Claici, Edward Chien, Farzaneh Mirzazadeh, Justin Solomon, Hierarchical Optimal Transport for Document Representation, NeurIPS 2019. (NeurIPS19_Paper)
Topics: Document embedding, Hierarchical optimal transport, structured representation.
- Charlie Frogner, Farzaneh Mirzazadeh, Justin Solomon, Learning Embeddings into Entropic Wasserstein Spaces, ICLR 2019. (ICLR19_Paper) (ICLR19_Poster)
Topics: Object2cloud embedding, increased embedding capacity, Wasserstein space.
- Farzaneh Mirzazadeh, Solving Association Problems with Convex Co-embedding, University of Alberta, 2016 (PhD Thesis)
Topics: Multi-modal embedding, assocaition learning, graph embedding
- Farzaneh Mirzazadeh, Siamak Ravanbakhsh, Nan Ding, and Dale Schuurmans, Embedding Inference for Structured Multilabel Prediction, Neural Information Processing Systems, 2015. (NIPS15-Paper, NIPS15_Poster)
Topics: Combination of symbolic reasoning and learning, object2ball embedding, logic embedding, multilabel classification, sructured prediction
- Farzaneh Mirzazadeh, Martha White, Andras Gyorgy, and Dale Schuurmans, Scalable metric learning for co-embedding, European Conference on Machine Learning , 2015, Oral (ECML15-Paper, ECML15-Talk Slides, ECML15-Poster)
Topics: Equivalence of global and locally optimal solutions.
- Farzaneh Mirzazadeh, Yuhong Guo and Dale Schuurmans, Convex co-embedding. In Twenty-Eighth Annual Conference on Artificial Intelligence, 2014, Oral (Paper, Talk Slides, Poster)
- Farzaneh Mirzazadeh and Dale Schuurmans, Data dependent loss functions for focused generalization and transfer learning, In ICML workshop, 2011 (ICMLWS11)
- Farzaneh Mirzazadeh, Using SNP data to predict radiation toxicity for prostate cancer patients, MSc thesis, University of Alberta, 2009.
- Farzaneh Mirzazadeh, and Saeed Bagheri Shouraki, Online recognition of handwritten Persian words using a novel hierarchical fuzzy system, (ICSCIS 07)
- Farzaneh Mirzazadeh, Online recognition of handwritten Persian words, MSc thesis, Sharif University of Technology, 2007
- Farzaneh Mirzazadeh, Babak Behsaz, and Hamid Beigy, A new learning algorithm for MAXQ hierarchical reinforcement learning method, in International Conference on Information and Communication Technology - (IEEE ICICT '07)
- Reza Safabakhsh and Farzaneh Mirzazadeh, AUT-Talk: a Farsi talking head, in Information and Communication Technologies, (IEEE ICTTA 06)
- Farzaneh Mirzazadeh, Design and implementation of a Farsi talking head (including a Farsi text to speech synthesis engine), BSc thesis, Amirkabir University of Technology, 2005.