https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354122
-- might be useful to disentangle representations
-- here it seems like they are removing the styles from potraits to match with face images (supratim?)
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354123
-- given a source input image and reference images with a target attribute, our goal is to generate a new image (i.e., target image) that not only possesses the new attribute but also keeps the same or similar content with the source image
-- idea can this network in figure 2 be used to train GAN to generate features with specific attributes ? I am not talking about the face images but suppose like text/image data features and you need to have this set of attributes... can it generate a set of attributes ? instead of a single attribute only ?
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354151
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354171
-- from the paper "... In many visual domains (like fashion, furniture, etc.) the search for products on online platforms requires matching textual queries to image content. For example, the user provides a search query in natural language (e.g.,pink floral top) and the results obtained are of a different modality (e.g., the set of images of pink floral tops). Recent work on multimodal representation learning enables such cross-modal matching by learning a common representation space for text and image. While such representations ensure that the n-dimensional representation of pink floral top is very close to representation of corresponding images, they do not ensure that the first k1 (< n) dimensions correspond to color, the next k2 (< n) correspond to style and so on. In other words, they learn entangled representations where each dimension does not correspond to a specific attribute. We propose two simple variants which can learn disentangled common representations for the fashion domain wherein each dimension would correspond to a specific attribute (color, style, silhoutte, etc.). ..."
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354192
-- divide the data into two random sets and using a clean monitoring set interplay the predictions between the two network to clean the noisy data
-- given for single modality, a heuristic algorithm
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354189
-- might be useful for few shot learning or zero-hot learning
-- from the paper "This strategy allows us to recognize unseen complex concepts from simple visual primitives. ..."
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354186
-- "... In existing works, the encoding-decoding nets and GANs are integrated by sharing weights on the generator/decoder, thus the two losses are backpropagated to the generator/decoder simultaneously, where a weighting factor is needed to balance the interaction between the two losses. The decoupled learning avoids the interaction and thus removes the requirement of the weighting factor, essentially improving the generalization capacity of the designed model to different applications ..."
-- "...The other contribution of the paper is the design of a new evaluation metric to measure the image quality of generative models. ..." -- might be useful to measure the effectiveness of our generative network
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354242
-- clean noisy labeled image data , a two stage framework where in the first stage, a small portion of the images from the noisy training set is identified of which the labels are correct with a high probability , in the second stage a deep neural network is trained in a semi - supervised manner
-- single modality, work, heuristic, and applicable on very noisy datasets, based on my understanding it is only applicable on images (they do some feaure augmentations which are not possible on normal features of data)
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354290
-- "... Given a query image, a list of attributes are created and returned to the user from the classification scores. After applying demonstrated attribute manipulations (denoted with green color), our proposed network finds the top-3 similar products, ranked from left to right. ..."
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8354191
-- they have re-defined the softmax loss and used a cosine loss instead
-- might be useful for my work to define a different probability model which can be used to model non-binary similarity between the data items
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8451239
-- "... The sheer amount of personal data being transmitted to cloud services and the ubiquity of cellphones cameras and various sensors, have provoked a privacy concern among many people. On the other hand, the recent phenomenal growth of deep learning that brings advancements in almost every aspect of human life is heavily dependent on the access to data, including sensitive images, medical records, etc. Therefore, there is a need for a mechanism that transforms sensitive data in such a way as to preserves the privacy of individuals, yet still be useful for deep learning algorithms. This paper proposes the use of Generative Adversarial Networks (GANs) as one such mechanism, and through experimental results, shows its efficacy .."
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8451372
-- the problem statement is interesting rather than the solution