Connect with us

Interviews

Jerry Xu, Co-Founder & CEO of Datatron – Interview Series

mm

Updated

 on

Jerry has extensive experience in machine learning, storage systems, online service, distributed systems, virtualization, and OS kernel. He has worked on high performance and large-scale systems at companies such as: Lyft, Box, Twitter, Zynga, and Microsoft. He has also authored the open-source project Lib Crunch and is a three-time Microsoft Gold Star Award winner. Jerry completed his master’s degree in computer science at Shanghai University. His most recent startup is Datatron.
Datatron began in 2016 after you left Lyft. How did you initially conceive of the Datatron business concept?

When we worked at Lyft, we noticed that data scientist usually comes from diverse background like Math, Physics, Bio-Engineering etc. It is often very hard for them to get the engineering part of how their models work although they have good intuition on the model and math. That motivated us to start Datatron. We are not trying to help data scientist to find the best algorithm. We only come into picture after the algorithm is decided and to make the model deployment, monitoring and management more efficient.

 

Datatron was selected by 500 Startups to be included in the 18th cohort of accelerator companies. How did this residency personally influence you, and how you manage Datatron?

We did learn a lot from StartX and 500 Startup experiences which includes how to pitch to investors, how to find product/market fit, how to run sales/marketing which we don’t have experience personally before.

 

Datatron is a management platform for ML, AI, and Data Science models. Could you elaborate on some of the functionalities that are offered by your platform?

Our product has four modules now, Model Deployment, Model Minoring, Model Challenger and Model Governance.

Model Deployment:

Create and scale model deployments in just a few clicks. Deploy models developed in any framework or language.

Model Monitoring:

Make better business decisions to save your team time and money. Monitor model performance and detect model decay as it happens.

Model Governance:

Spend less time on model validation, bias detection, and internal audit processes. Go from model development to internal auditing to production faster than ever.

 

One of the use cases of Datatron is Demand Forecasting which is important for enterprises which need to plan and allocate resources. How does machine learning play into this?

Demand usually changes with both seasonality and trend, which is a typical machine learning problem. Machine learning models like ARIMA, Recurrent Neural Network (RNN) can learn from historic data to find the trend and seasonality automatically and make predictions based on that.

 

Which framework models (for example, TensorFlow) do you currently support?

We support most of the popular machine learning frameworks like sklearn, TensorFlow, H2O, R, SAS etc.

 

Which languages do models need to be built in to be supported by Datatron?

We support models in their native languages – Python, R, Java etc.

 

What are some of the types of industries which are best served by using the Datatron platform?

Fundamentally our platform is a horizontal solution which means it can be used by lots of different industries. As of now, we are trying to focus on financial service and telecommunication.

 

What are some of the most challenging aspects of data science that companies face, and why does Datatron solve this for them?

Lots of companies have different data science team already and those teams are using different tools to build their model and have different practice to manage their models. More and more enterprise realized that model is becoming an asset and will impact their top line directly. Having a platform can standardize the machine learning practice across the company becomes critical and required. Our platform can help to solve those issues.

 

Is there anything else that you would like to share about Datatron?

We got lots of inbound interests from big enterprises. At the same time, we are also building up our sales and marketing team to reach out to potential customers actively.

To learn more visit Datatron.

Spread the love

Antoine Tardif is a Futurist who is passionate about the future of AI and robotics. He is the CEO of BlockVentures.com, and has invested in over 50 AI & blockchain projects. He is the Co-Founder of Securities.io a news website focusing on digital securities, and is a founding partner of unite.AI. He is also a member of the Forbes Technology Council.

Computing

Artificial Intelligence Enhances Speed of Discoveries For Particle Physics

mm

Updated

 on

Researchers at MIT have recently demonstrated that utilizing artificial intelligence to simulate aspects of particles and nuclear physics theories can lead to faster algorithms, and therefore faster discoveries when it comes to theoretical physics. The MIT research team combined theoretical physics with AI models to accelerate the creation of samples that simulate interactions between neutrons, protons, and nuclei.

There are four fundamental forces that govern the universe: gravity, electromagnetism, the weak force, and the strong force. The strong, weak, and electromagnetic forces are studied through particle physics. The traditional method of studying particle interactions requires running numerical simulations of these interactions between particles, typically taking place at 1/10th or 1/100th the size of a proton. These studies can take a long time to complete due to limited computing power, and there are many problems that physicists know how to tackle in theory yet cannot address to said computational limitations.

MIT Physics professor Phiala Shanahan is the head of a research group that uses machine learning models to create new algorithms that can speed up particle physics studies. The symmetries found within physics theories (features of the physical system that stay constant even as conditions change) can be incorporated into machine learning algorithms to produce algorithms more suited to particle physics studies. Shanahan explained that the machine learning models aren’t being used to process large amounts of data, rather they are being used to integrate particle symmetries, and the inclusion of these attributes within a model means that computations can be done more quickly.

The research project was lead by Shanahan and it includes several members of the Theoretical Physics team at NYU, as well as machine-learning researchers from Google DeepMind. The recent study is just one of a series of ongoing and recently completed studies aimed at leveraging the power of machine learning to solve theoretical physics problems that are currently impossible with modern computation schemas. According to MIT graduate student Gurtej Kanwar, the problems that the machine-learning boosted algorithms are trying to solve will help scientists understand more about particle physics, and they are useful in making comparisons against results derived by large-scale particle physics experiments (like those conducted at CERN’s Large Hadron Collider). By comparing the results of the large-scale experiments with the AI algorithms, scientists can get a better idea of how their physics models should be constrained, and when those models break down.

Currently, the only method that scientists can reliably use to investigate the Standard Model of particle physics is one where samples/snapshots are taken of fluctuations occurring in a vacuum. Researchers can gain insight into the properties of the particles and what happens when those particles collide. However, taking samples like this is expensive and it is hoped that AI techniques can make taking samples a cheaper, more efficient process. The snapshots taken of the vacuum can be used much like image training data in a computer vision AI model. The quantum snapshots are used to train a model that can create samples in a much more efficient manner, accomplished by taking samples in an easy-to-sample space and running the samples through the trained model.

The research has created a framework intended to streamline the process of creating machine-learning models based on physics symmetries. The framework has already been applied to simpler physics problems and the research team is now attempting to scale up their approach to work with cutting edge calculations. As Kanwar explained via Phys.org:

“I think we have shown over the past year that there is a lot of promise in combining physics knowledge with machine learning techniques. We are actively thinking about how to tackle the remaining barriers in the way of performing full-scale simulations using our approach. I hope to see the first application of these methods to calculations at scale in the next couple of years.”

Spread the love
Continue Reading

AI 101

What Is Synthetic Data?

mm

Updated

 on

What is Synthetic Data?

Synthetic data is a quickly expanding trend and emerging tool in the field of data science. What is synthetic data exactly? The short answer is that synthetic data is comprised of data that isn’t based on any real-world phenomena or events, rather it’s generated via a computer program. Yet why is synthetic data becoming so important for data science? How is synthetic data created? Let’s explore the answers to these questions.

What is a Synthetic Dataset?

As the term “synthetic” suggests, synthetic datasets are generated through computer programs, instead of being composed through the documentation of real-world events. The primary purpose of a synthetic dataset is to be versatile and robust enough to be useful for the training of machine learning models.

In order to be useful for a machine learning classifier, the synthetic data should have certain properties. While the data can be categorical, binary, or numerical, the length of the dataset should be arbitrary and the data should be randomly generated. The random processes used to generate the data should be controllable and based on various statistical distributions. Random noise may also be placed in the dataset.

If the synthetic data is being used for a classification algorithm, the amount of class separation should be customizable, in order that the classification problem can be made easier or harder according to the problem’s requirements. Meanwhile, for a regression task, non-linear generative processes can be employed to generate the data.

Why Use Synthetic Data?

As machine learning frameworks like TensorfFlow and PyTorch become easier to use and pre-designed models for computer vision and natural language processing become more ubiquitous and powerful, the primary problem that data scientists must face is the collection and handling of data. Companies often have difficulty acquiring large amounts of data to train an accurate model within a given time frame. Hand-labeling data is a costly, slow way to acquire data. However, generating and using synthetic data can help data scientists and companies overcome these hurdles and develop reliable machine learning models a quicker fashion.

There are a number of advantages to using synthetic data. The most obvious way that the use of synthetic data benefits data science is that it reduces the need to capture data from real-world events, and for this reason it becomes possible to generate data and construct a dataset much more quickly than a dataset dependent on real-world events. This means that large volumes of data can be produced in a short timeframe. This is especially true for events that rarely occur, as if an event rarely happens in the wild, more data can be mocked up from some genuine data samples. Beyond that, the data can be automatically labeled as it is generated, drastically reducing the amount of time needed to label data.

Synthetic data can also be useful to gain training data for edge cases, which are instances that may occur infrequently but are critical for the success of your AI. Edge cases are events that are very similar to the primary target of an AI but differ in important ways. For instance, objects that are only partially in view could be considered edge cases when designing an image classifier.

Finally, synthetic datasets can minimize privacy concerns. Attempts to anonymize data can be ineffective, as even if sensitive/identifying variables are removed from the dataset, other variables can act as identifiers when they are combined. This isn’t an issue with synthetic data, as it was never based on a real person, or real event, in the first place.

Uses Cases for Synthetic Data

Synthetic data has a wide variety of uses, as it can be applied to just about any machine learning task. Common use cases for synthetic data include self-driving vehicles, security, robotics, fraud protection, and healthcare.

One of the initial use cases for synthetic data was self-driving cars, as synthetic data is used to create training data for cars in conditions where getting real, on-the-road training data is difficult or dangerous. Synthetic data is also useful for the creation of data used to train image recognition systems, like surveillance systems, much more efficiently than manually collecting and labeling a bunch of training data. Robotics systems can be slow to train and develop with traditional data collection and training methods. Synthetic data allows robotics companies to test and engineer robotics systems through simulations. Fraud protection systems can benefit from synthetic data, and new fraud detection methods can be trained and tested with data that is constantly new when synthetic data is used. In the healthcare field, synthetic data can be used to design health classifiers that are accurate, yet preserve people’s privacy, as the data won’t be based on real people.

Synthetic Data Challenges

While the use of synthetic data brings many advantages with it, it also brings many challenges.

When synthetic data is created, it often lacks outliers. Outliers occur in data naturally, and while often dropped from training datasets, their existence may be necessary to train truly reliable machine learning models. Beyond this, the quality of synthetic data can be highly variable. Synthetic data is often generated with an input, or seed, data, and therefore the quality of the data can be dependent on the quality of the input data. If the data used to generate the synthetic data is biased, the generated data can perpetuate that bias. Synthetic data also requires some form of output/quality control. It needs to be checked against human-annotated data, or otherwise authentic data is some form.

How Is Synthetic Data Created?

Synthetic data is created programmatically with machine learning techniques. Classical machine learning techniques like decision trees can be used, as can deep learning techniques. The requirements for the synthetic data will influence what type of algorithm is used to generate the data. Decision trees and similar machine learning models let companies create non-classical, multi-modal data distributions, trained on examples of real-world data. Generating data with these algorithms will provide data that is highly correlated with the original training data. For instances where the typical distribution of data is known , a company can generate synthetic data through use of a Monte Carlo method.

Deep learning-based methods of generating synthetic data typically make use of either a variational autoencoder (VAE) or a generative adversarial network (GAN). VAEs are unsupervised machine learning models that make use of encoders and decoders. The encoder portion of a VAE is responsible for compressing the data down into a simpler, compact version of the original dataset, which the decoder then analyzes and uses to generate an a representation of the base data. A VAE is trained with the goal of having an optimal relationship between the input data and output, one where both input data and output data are extremely similar.

When it comes to GAN models, they are called “adversarial” networks due to the fact that GANs are actually two networks that compete with each other. The generator is responsible for generating synthetic data, while the second network (the discriminator) operates by comparing the generated data with a real dataset and tries to determine which data is fake. When the discriminator catches fake data, the generator is notified of this and it makes changes to try and get a new batch of data by the discriminator. In turn, the discriminator becomes better and better at detecting fakes. The two networks are trained against each other, with fakes becoming more lifelike all the time.

Spread the love
Continue Reading

Commerce

The Science of Real-Estate: Matching and Buying

mm

Updated

 on

Your data knows you best, let it find your dream home. The real-estate industry sits on tons of data that goes unused every year. In this article, we discuss how advanced technologies are helping real-estate investors, brokers, and companies utilize the mass amount of information within the industry to help people find their dream homes.

In 2017, a Field Actions Science Reports article addresses the impact of AI, machine learning, and predictive analytics on the real-estate sector:

“The practice of AI-powered Urban Analytics is taking off within the real-estate industry. Data science and algorithmic logic are close to the forefront of new urban development practices. How close? is the question — experts predict that digitization will go far beyond intelligent building management systems. New analytical tools with predictive capabilities will dramatically affect the future of urban development, reshaping the real-estate industry in the process.”

Fast forward to 2020: leaving hype traps behind, we acknowledge the transformative effects of data literacy, digitalization strategies, and technology advancements. Predictive analytics, machine learning, and AI-powered applications are still leading innovation in a variety of industries, well beyond the real-estate sector. From the most boring ML applications to the most interesting NLP & OCR automation efforts, industry leaders have learned to leverage these powerful tools to their advantage.

Today we catch up with 3 real-estate use cases. They are meant to illustrate how modern software stacks and intuitive interfaces interplay with Machine Learning and data engineering to create unique products and services.

science of real estate one

science of real estate: Your data knows you best, let it find you the perfect home.

Home buying processes

Today’s real-estate market poses an interesting machine learning challenge: is there a formula for matching the right home-buyers with the right properties at the right prices? Seeking to build accurate home matching and discovery services is what keeps researchers and industry professionals on their toes. With huge data volumes available to them, and inspired by high accuracy of online recommender systems (Netflix, anyone?), home matching engines are seeing constant development, even in the not-so-technically-inclined real-estate sector. 

Orchard is a broker that leverages modern tech tools to improve home discovery services. By using machine learning algorithms, they come up with an answer to the most pressing question that home buyers ask: “What does my dream house look like?”. Additionally, algorithms may help them answer a follow-up question: “Which compromises are I (not) willing to make?”. 


Co-Founder and Chief Product & Marketing Officer, Phil DeGisi clarifies:

Home Match is the first-ever home search algorithm that lets people choose the features that matter most to them. We ask buyers a series of questions about what they value and consider “must-haves” and “nice to haves” in a home – such as a kitchen island, pool in the backyard, and commute time within seconds. Orchard assigns a personal match score to every home in the search area.

Like this, the buyers are matched to legitimate house buying opportunities and the entire process becomes easier for all parties involved. 

Users of house matching systems get to enjoy an experience characterized by increased personalization and usability. Search results are ranked according to their profiles and easy-to-use, interactive interfaces replace plain old real-estate catalogs.

“Orchard has also developed another industry-first, Photo Switch, which takes these personalized search results and displays them in a more visually useful and personalized way. To do this, Orchard created a machine-learning model to scan photos of every home on the market and determine which rooms are in each photo. This feature is the first of its kind and lets users easily compare their “must-haves” all at once. Whether it’s a chef’s kitchen, a fenced-in backyard, or a cozy living room, home-buyers can now view each room side-by-side in one browser, with the click of a single button.”

Such functionality is only possible due to the seamless interplay of modern tech tools. Web platforms, virtual reality SDKs, image processing algorithms as well as machine learning frameworks all contribute to create a unique real-estate experience.

Commercial real-estate valuations

Another crucial step in commercial real-estate is property valuation. Automated Valuation Models are as old as the industry itself, given the task of evaluating properties and establishing pricing schemes. Traditionally, these models were mostly based on historical sales data. However, models relying on past behavior only are missing out on a lot of other data sources.

Predictive analytics and modern data collection infrastructures are built to integrate external data sources and train algorithms based on heterogeneous data types. Instead of using a single data type that offers a limited perspective on a property, unified data architectures offer a 360-degree view and integrate external data sources: market demand, macroeconomic data, rental values, capital markets, jobs, traffic, etc. Since there are no hard limits to the data that can be used by a property valuation model, predictive analytics is a powerful tool available to real-estate agencies. 

Smart Capital offers such a modern solution to property valuation. They use predictive analytics for the valuation of real-estate properties and promise to deliver a full report within one business day. Their CEO, Laura Krashakova, offers some insights into how they achieve this.

The technology enables data processing and property valuation in real-time and gives individuals access to data previously available only to local brokers. Local insights such as the popularity of the location, amenities in the area, quality of public transport, proximity to major highways, and foot traffic are now readily available and are scored for ease of comparison.

There are two aspects that make such a service possible in the first place: the ease of access and the possibility to deliver real-time insights. Mobile & web platforms make it easy for customers to access, upload, and visualize their data, regardless of their location. All that is needed is an internet connection. At the same time, predictive analytics frameworks are crunching data in real-time, at the speed of milliseconds. Once new data events occur, they are collected and included in the latest analysis report. No need to wait for time-consuming, intensive computations, since all of that computation can now happen almost instantly, in the cloud.

Once again, the interplay of modern technologies makes it possible to offer a seamless experience based on real-time insights. At the same time, the variety of external data sources becomes a guarantee for increased valuation accuracy. This saves time, money, and headaches for all parties involved.

Streamlined loan application processes

Another commercial real-estate process that poses an interesting challenge is the loan application. A challenge not only for the confused homebuyers but for machine learning models as well. Credit approval models need access to all kinds of data, from personal information, to credit history, historical transactions, and employment history. Manually identifying and integrating all these data sources can quickly turn into a tedious, time-consuming, and annoying task. Moreover, manual processing comes with a high risk of erroneous entries throughout the application. These aspects have turned the manual loan application process into a bottleneck for real-estate transactions.

If only some automated solution existed to take some of the pain away…

Beeline is a company focused on streamlining the loan application process. Their intuitive mobile interface guides buyers through loan applications in minutes. The entire process takes only 15 minutes and claims to save home buyers a lot of headaches. The way they do this is incredibly simple: their service connects to a variety of personal data sources (such as the bank, pay and tax info), uses natural language processing(NLP) to read and collect info, integrates and analyzes all the data in real-time. Like this, tedious and time-consuming processes are bypassed and home-buyers can enjoy streamlined loan application processes.

How is that possible, you’re wondering? 

Their service is only possible by integrating a mobile-first experience, intelligent processing capabilities, as well as state of the art user design. Their loan guide is delivered via a chat interface, which gives the users an easy way to find answers to their questions. NLP algorithms are backing these interactions and help create a personalized experience.

At the same time, automated evaluation algorithms happen in the background, just as the buyer is filling in forms. This shows how automation is key to the success of their service. And the seamless interplay of tech tools is what makes this automation possible in the first place.

What’s next?

A powerful mix of tech trends is at the forefront of real-estate innovation: increased data availability, advancements in data processing capabilities, and the ubiquity of machine learning algorithms. They all make it possible to tackle the most challenging applications, in an intelligent, automated, and error-free manner. 

On top of that, cloud computing capabilities and modern storage architectures make it possible to extract insights from data in real-time, build complex predictive models, and integrate a variety of data sources. All this makes it possible to foresee the future, innovate, and keep a competitive advantage.

image sources: Canva

Spread the love
Continue Reading