Sr. Lead Data Sciences
Envestnet | Yodlee
Member since 3 years
I am a Data Scientist at Envestnet Yodlee. I work in deploying Deep Learning, Reinforcement Learning, and Semi-Supervised learning based products.
I completed my BTech from IIIT Delhi, working on machine learning and distributed systems. I completed my MTech degree from IIT Bombay. At IIT Bombay, the majority of my time was spent in developing deep learning, reinforcement learning, and computer vision software. I worked on a lot of cool AI - agents for self-driving cars, robot soccer, Atari games, and Carrom. I also interned in Scilab, where I reverse engineered functions from the Matlab Image Processing Toolbox. As a part of my MTech thesis, I developed a Multi-Armed-Bandit framework for autonomous agents to supervise their own training.
Algorithms that learn to solve tasks by watching (one) Youtube videoSamiran RoySr. Lead Data SciencesEnvestnet | YodleeShibsankar DasSr. Lead Data ScientistEnvestnet | Yodlee
schedule 2 years agoSold Out!
Imitation Learning has been the backbone of Robots learning from demonstrator's behavior. Join us to know more about How to train a robot to perform task like acrobatics etc.
Two branches of AI - Deep Learning, and Reinforcement Learning are now responsible for many real-world applications. Machine Translation, Speech Recognition, Object Detection, Robot Control, and Drug Discovery - are some of the numerous examples.
Both approaches are data-hungry - DL requires many examples of each class, and RL needs to play through many episodes to learn a policy. Contrast this to human intelligence. A small child can typically see an image just once, and instantly recognize it in other contexts and environments. We seem to possess an innate model/representation of how the world works, which helps us grasp new concepts and adapt to new situations fast. Humans are excellent one/few shot learners. We are able to learn complex tasks by observing and imitating other humans (eg: cooking, dancing or playing soccer) - despite having a different point of view, sense modalities, body structure, mental facility.
Humans may be very good at picking up novel tasks, but Deep RL agents surpass us in performance. Once a Deep RL has learned a good representation , it is easy to surpass human performance in complex tasks like Go, Dota 2, and Starcraft. We are biologically limited by time, memory and computation (A computer can be made to simulate thousands of plays in a minute).
RL struggles with tasks that have sparse rewards. Take an example of a soccer playing robot - controlled by applying a torque to each one of its joints. The environment rewards you when it scores a goal. If the policy is initialized randomly (we apply a random torque to each joint, every few milliseconds) the probability of the robot scoring a goal is negligible - it won't even be able to learn how to stand up. In tasks requiring long term planning or low-level skills, getting to that initial reward can prove impossible. These situations have the potential to greatly benefit from a demonstration - in this case showing the robot how to walk and kick - and then letting it figure out how to score a goal.
We have an abundance of visual data on humans performing various tasks, in the public domain, in the form of videos from sources like YouTube. In Youtube alone, 400 hours of videos are uploaded every minute, and it is easy to find demonstration videos for any skill imaginable. What if we could harness this by designing agents that could learn how to perform tasks - just by watching a video clip?
Imitation Learning, also known as apprenticeship learning, teaches an agent a sequence of decisions through demonstration, often by a human expert. It has been used in many applications such as teaching drones how to fly and autonomous cars how to drive - It relies on domain engineered features - or extremely precise representations such as mocap . Directly applying imitation learning to learn from videos proves challenging, there is a misalignment of representation between the demonstrations and the agent’s environment. For example: How can a robot sensing its world through a 3d point cloud - learn from a noisy 2d video clip of a soccer player dribbling?
Leveraging recent advances in Reinforcement Learning, Self Supervised Learning and Imitation Learning   , We present a technical deep dive into an end to end framework which:
1) Has prior knowledge about the world intelligence through Self-Supervised Learning - A relatively new area which seeks to build efficient deep learning representations from unlabelled data but training on a surrogate task. The surrogate task can be rotating an image and predicting the rotation angle or cropping two patches of the image, and predicting their relative tasks - or a combination of several such objectives.
2) Has the ability to align the representation of how it senses the world, with that of the video - allowing it to learn diverse tasks from video clips.
3) Has the ability to reproduce a skill, from only a single demonstration - using applied techniques from imitation learning
Semi-Supervised Insight generation from petabyte scale Text dataSamiran RoySr. Lead Data SciencesEnvestnet | YodleeShibsankar DasSr. Lead Data ScientistEnvestnet | Yodlee
schedule 2 years agoSold Out!
Existing state-of-the-art supervised methods in Machine Learning require large amounts of annotated data to achieve good performance and generalization. However, manually constructing such a training data set with sentiment labels is a labor-intensive and time-consuming task. With the proliferation of data acquisition in domains such as images, text and video, the rate at which we acquire data is greater than the rate at which we can label them. Techniques that reduce the amount of labelled data needed to achieve competitive accuracies are of paramount importance for deploying scalable, data-driven, real-world solutions. Semi-Supervised Learning algorithms generally provide a way of learning about the structure of the data from the unlabelled examples, alleviating the need for labels.
At Envestnet | Yodlee, we have deployed several advanced state-of-the-art Machine Learning solutions which process millions of data points on a daily basis with very stringent service level commitments. A key aspect of our Natural Language Processing solutions is Semi-supervised learning (SSL): A family of methods that also make use of unlabelled data for training – typically a small amount of labelled data with a large amount of unlabelled data. Pure supervised solutions fail to exploit the rich syntactic structure of the unlabelled data to improve decision boundaries.
There is an abundance published work in the field - but few papers have succeeded in showing significantly better results than state-of-the-art supervised learning. Often, methods have simplifying assumptions that fail to transfer to real-world scenarios. There is a lack of practical guidelines for deploying effective SSL solutions. We attempt to bridge that gap by sharing our learning from successful SSL models deployed in production.
We will talk about best practices and challenges in deploying SSL solutions in NLP - We shall cover:
- Our findings while working on SSL.
- Techniques which have worked for us, and which have not
- Which SSL method is suitable to solve a given use-case.
- How to deal with different distributions for labelled and unlabelled data
- How to quantify the effectiveness of each point in our training data
- How to build a feedback loop that chooses points for training that result in the greatest accuracy boosts and
- The effect of relative sizes of labelled and unlabelled data
Reinforcement Learning: Demystifying the hype to successful enterprise applicationsSamiran RoySr. Lead Data SciencesEnvestnet | YodleeDr. Om DeshmukhSenior Director - Data Science & InnovationYodlee
schedule 3 years agoSold Out!
In 2014, Google accquired DeepMind, a small, london-based AI startup for $500 million. DeepMind was conducting research on AI that would learn to play computer games in a fashion similar to humans. In 2015, Deepmind published a paper in Nature, describing a learning algorithm called Deep-Q-Learning which was able to achieve superhuman performance on a diverse range of Atari 2600 games. They achieved this without any domain specific engineering - The algorithm took only the raw game images as input, and was guided by the game score. Believed by many to be the first steps in Artificial General Intelligence, DeepMind achieved this by pioneering the fusion of two fields of research - Reinforcement Learning(RL) and Deep Learning.
RL is a learning paradigm inspired by operant conditioning which closely mimics the human learning process. It shifts focus from ML based pattern recognition solutions to learning through trial and error via interaction with an environment, guided by a reward signal or reinforcement. Imagine an agent teaching itself how to steer by navigating the streets of Grand Theft Auto - and transferring this knowledge to a driverless car. Think of team of autonomous robots collaborating to outwit their opponents in a game of Robot Soccer. Any practical real-world application suffers from the curse of dimensionality (A camera mounted on a robot feeding it a 64*64 grayscale image will have 256^(4096) input possibilities). A Deep Neural Network automatically learns compact and efficient feature representations from noisy, high-dimensional sensory inputs in its hidden layers, giving RL algorithms the edge to scale up and give incredible results in dynamic and complex domains.
The most notable example of this is AlphaGo Zero - the latest version of AlphaGo, the first computer program to defeat a world champion at the game of Go (Also called Chinese Checkers). AlphaGo Zero uses RL to learn by playing games against itself, starting from completely random play, and quickly surpasses human expert performance. Not only is the game extremely complex (A 19*19 Go board can represent 10^170 states of play), accomplished Go players often struggle to evaluate whether a certain move is good or bad. Most AI researchers were astonished by this feat, as it was speculated that it would take atleast a decade for a computer to play Go at an expert human level.
RL, which was largely confined to academia for several decades is now beginning to see some successful applications and products in the industry, in fields such as robotics, automated trading systems, manufacturing, energy, dialog systems and recommendation engines. For most companies, it is an exciting prospect due to the AI hype, but very few organizations have identified use cases where RL may play a valuable role. In reality, RL is best suited for a niche class of problems where it can help automate some tasks(or augment a human expert). The focus of this presentation will be to give a practical introduction to the RL Setting, how to formulate problems into RL, and presenting successful use cases in the industry.
No more submissions exist.
No more submissions exist.