AI News, Square's Machine Learning Infrastructure and Applications

Square's Machine Learning Infrastructure and Applications

Machine learning applications like fraud detection and recommendation have played a key role in helping Square achieve their mission to rethink buying and selling.

In this talk, Dr. Rong Yan (Director of Data Science and Infrastructure, Square), gives a high-level overview of data applications at Square followed by a deep dive on how machine learning is used in our industrial leading fraud detection models.

Machine Learning for Fraud Detection – Modern Applications and Risks

By using machine learning to identifying your company’s own biggest fraud risks and predict and guard against those risks, you can protect your company, your clients and your reputation, while cutting operational costs and increasing user confidence.

Kevin Lee: Ten or fifteen years ago mainly people were concerned about payment fraud or credit card fraud online…It has become a bit more pervasive, where the issue is not just about the credit card, it’s about identities and even the way people interact online.

Now eCommerce has come up…originally eBay was a master merchant but then they enabled people to sell stuff from their garage and that introduced a whole new variable where before, if you were a merchant, you obviously had confidence in yourself, of course I’m going to deliver these goods that I’m selling, but now you get into a scenario where I could be bad, the merchant could be bad or both could be bad and so it becomes more complex to figure out who’s bad, what’s the story and what is going on in this space.

Now what’s happening, partly because of data breached that are occurring, another reason is that people are just moving around more of their identities online, account takeover is becoming the next thing…Many vendors out there have done a very good job of spotting and killing fake accounts and so the fraudsters and spammers out there, they’re running a business as well, so if they’re not getting their return on investment, by creating these fake accounts, they need to go elsewhere…The next step is around compromising identities and accounts…we are seeing much more targeted and specific hacking…The incidences may be less, but the damage per incidence is much more.

Kevin Lee: It can vary per company a bit but really there are two main buckets, there’s the data security side of it, in terms of theft of credit cards are they hashed, or encrypted and then there’s the info security in terms of risk and abuse of payment, promotion or content…  trust and safety and risk goes beyond just financial.

Fraud now it sounds like extends to something like someone on Yelp doing fake reviews, or someone  with a blogging service with some really spammy, scraped together, re-hashed articles that are going all over the place…for these example we could use the term fraudulent content, which clearly will have, at a certain scale, a negative financial impact on a company.

(16.29) I guess companies now are more aware of this type of abuse and the potential financial implications…If I’m detecting a fraudulent charge, if I’m detecting an unusual eBay customer profile, if I’m detecting some unusual review behaviors on Yelp, whether this is a market place and it’s content, whether it’s engagement, whether it’s a transaction…these are all examples of I guess what we could call anomaly detection.

briefings - August 8 9

Age old exploits give adversaries administrator level privileges without physical access to the machine and to make matters worse, the remote desktop protocol is enabled by default on each and every machine.

All of this is well-known and well-documented, however there are lessons to be learned that go beyond hacking, lessons that effect society as a whole.The single most important concern of any electoral process is the trust of the voters: winners and losers alike must be convinced of the quality of the electoral process so that all are able to accept the outcome.

This is a tall order, because, as we all know by now, national elections use election technologies in highly contested adversarial environments, where network, hardware, software, and configuration processes must be assumed to be under the adversary's control.

However, we could clearly establish that some WinVote voting machines were used for purposes other than voting: One Voting machine was used to rip songs from CDs and broadcast MP3s, most notably, perhaps, a Chinese song from 1995: 白雪-千古绝唱.mp3.Trust in elections cannot be achieved through technology alone - it can only be achieved by the means of producing evidence and checking it for consistency.

Age old exploits give adversaries administrator level privileges without physical access to the machine and to make matters worse, the remote desktop protocol is enabled by default on each and every machine.

The single most important concern of any electoral process is the trust of the voters: winners and losers alike must be convinced of the quality of the electoral process so that all are able to accept the outcome.

This is a tall order, because, as we all know by now, national elections use election technologies in highly contested adversarial environments, where network, hardware, software, and configuration processes must be assumed to be under the adversary's control.

The Enterprise Immune System

Modeled on the human immune system, Darktrace’s technology is the world’s leading, enterprise-grade AI with over 7,000 deployments worldwide.

Machine Learning is Fun! Part 3: Deep Learning and Convolutional Neural Networks

First, the good news is that our “8” recognizer really does work well on simple images where the letter is right in the middle of the image: But now the really bad news: Our “8” recognizer totally fails to work when the letter isn’t perfectly centered in the image.

We can just write a script to generate new images with the “8”s in all kinds of different positions in the image: Using this technique, we can easily create an endless supply of training data.

But once we figured out how to use 3d graphics cards (which were designed to do matrix multiplication really fast) instead of normal computer processors, working with large neural networks suddenly became practical.

It doesn’t make sense to train a network to recognize an “8” at the top of a picture separately from training it to recognize an “8” at the bottom of a picture as if those were two totally different objects.

Instead of feeding entire images into our neural network as one grid of numbers, we’re going to do something a lot smarter that takes advantage of the idea that an object is the same no matter where it appears in a picture.

Here’s how it’s going to work, step by step — Similar to our sliding window search above, let’s pass a sliding window over the entire original image and save each result as a separate, tiny picture tile: By doing this, we turned our original image into 77 equally-sized tiny image tiles.

We’ll do the exact same thing here, but we’ll do it for each individual image tile: However, there’s one big twist: We’ll keep the same neural network weights for every single tile in the same original image.

It looks like this: In other words, we’ve started with a large image and we ended with a slightly smaller array that records which sections of our original image were the most interesting.

We’ll just look at each 2x2 square of the array and keep the biggest number: The idea here is that if we found something interesting in any of the four input tiles that makes up each 2x2 grid square, we’ll just keep the most interesting bit.

So from start to finish, our whole five-step pipeline looks like this: Our image processing pipeline is a series of steps: convolution, max-pooling, and finally a fully-connected network.

For example, the first convolution step might learn to recognize sharp edges, the second convolution step might recognize beaks using it’s knowledge of sharp edges, the third step might recognize entire birds using it’s knowledge of beaks, etc.

Here’s what a more realistic deep convolutional network (like you would find in a research paper) looks like: In this case, they start a 224 x 224 pixel image, apply convolution and max pooling twice, apply convolution 3 more times, apply max pooling and then have two fully-connected layers.

Contact WooCommerce for Square for WooCommerce support

Get the Square extension for WooCommerce at no cost and instantly start accepting payments.

With Square and WooCommerce, you can track your inventory and sales across all channels in one place.

As with any transaction, there’s always a risk that your customer will dispute the validity of the purchase.

Our fraud prevention team watches over your account’s security 24/7.

We take a holistic look at the entire Square ecosystem to analyze a wide variety of signals and keep ahead of fraud trends.

We’re the merchant of record for every transaction, which means we’re dedicated to keeping your business safe.

Anomaly Detection: Algorithms, Explanations, Applications

Anomaly detection is important for data cleaning, cybersecurity, and robust AI systems. This talk will review recent work in our group on (a) benchmarking ...

But what *is* a Neural Network? | Deep learning, chapter 1

Subscribe to stay notified about new videos: Support more videos like this on Patreon: Or don'

Lesson 2: Deep Learning 2018

NB: Please go to to view this video since there is important updated information there. If you have questions, use the forums at ..

Lecture 3 | Loss Functions and Optimization

Lecture 3 continues our discussion of linear classifiers. We introduce the idea of a loss function to quantify our unhappiness with a model's predictions, and ...

Build a Neural Net in 4 Minutes

How does a Neural network work? Its the basis of deep learning and the reason why image recognition, chatbots, self driving cars, and language translation ...

Customer Successes with Machine Learning (Google Cloud Next '17)

TensorFlow is rapidly democratizing machine intelligence. Combined with the Google Cloud Machine Learning platform, TensorFlow now allows any developer ...

Pedro Domingos: "The Master Algorithm" | Talks at Google

Machine learning is the automation of discovery, and it is responsible for making our smartphones work, helping Netflix suggest movies for us to watch, and ...

Azure Machine Learning: Predict Who Survives the Titanic - Jennifer Marsman - Duo Tech Talk

Interested in doing machine learning in the cloud? In this demo-heavy talk, I will set the stage with some information on the different types of machine learning ...

Building the Connected Enterprise with Machine Learning and Neo4j — Tim Ward, CluedIn

It goes without saying, every single company is wanting to use the Graph. Every single company is wanting to use Machine Learning. Not many companies are ...