Learn How to Perform Feature Extraction from Graphs using DeepWalk

• Extracting features from tabular or image data is a well-known concept – but what about graph data?
• Learn how to extract features from a graph using DeepWalk, a robust and scalable method
• We will also implement DeepWalk in Python to find similar Wikipedia pages

Creating Autonomous Entities: Optimizing Systems of Subsystems

I continue to be infatuated with the potential of autonomous entities that self-monitor, self-diagnose, self-fix and self-learn without human intervention. The vision of leveraging new digital technologies, growing reams of customer, product and operational data, and advanced analytics to create autonomous entities – farms, oil fields, factories, airports, theme parks, vehicles, ships, trains, etc. – seems like science fiction.

A First Look at Confidence Distributions

Using a probability distribution to characterize uncertainty is at the core of statistical inference. So, it seems natural to try to summarize the information about the parameters in statistical models with probability distributions. R. A. Fisher thought so. In fact, he expended a great deal of effort over more than thirty years, and put his professional reputation on the line trying to do so, with only limited success. Fisher’s central difficulty was that, in the Frequentist tradition to which he was committed, parameters are not random variables. They are fixed and immutable constituents of the statistical models describing the behavior of populations, which we must estimate because we generally only have access to samples from populations, not to the full populations themselves. Now Bayesians, of course, characterize parameters with probability distributions from the get-go. Parameters are given prior distributions and combined with the likelihood function generated by the data to produce posterior distributions that characterize the parameters. Fisher wanted the posterior distributions without having to assume the priors. This was a key motivating idea for his work on Fiducial probability.

How Randomness Can Arise From Determinism

Playing with a simple bean machine illustrates how deterministic laws can produce probabilistic, random-seeming behavior.


State-of-the-art Natural Language Processing for TensorFlow 2.0 and PyTorch. Transformers (formerly known as pytorch-transformers and pytorch-pretrained-bert) provides state-of-the-art general-purpose architectures (BERT, GPT-2, RoBERTa, XLM, DistilBert, XLNet, CTRL…) for Natural Language Understanding (NLU) and Natural Language Generation (NLG) with over 32+ pretrained models in 100+ languages and deep interoperability between TensorFlow 2.0 and PyTorch.

Find thy hyper-parameters for scikit-learn pipelines using Microsoft NNI

The goal of any machine learning algorithm is to find patterns in the sample data and then use those patterns to predict the outcomes given unseen sample (test) data. The patterns that a machine learning algorithm learns are encoded in the weights (also called parameters) of a model. However, a machine learning algorithm can be applied to different types of data and even for scenarios where the ‘type’ of data is same the distribution of the data could be different.

Human Bias in Machine Learning

Bias is an inescapable part of human nature. Previous research suggests that cognitive biases form to optimize brain function when humans are distracted. Biases are influenced by your environment, experiences, and can be difficult to eliminate. One such solution to minimize the effects of biases is to be aware of possible biases that you have or may encounter. In this post, I will be describing some of the impacts that different types of biases can cause in machine learning projects. With examples on the roots of the issues caused by these biases as well as reasoning on why bias can be useful.

Add Image Recognition to your Chatbot with Google Dialogflow and Vision API

Conversational AI use cases are diverse. They include customer support, e-commerce, controlling IoT devices, enterprise productivity and much more. In very simplistic terms, these use cases involve a user asking a specific question (intent) and the conversational experience (or the chatbot) responding to the question by making calls to a backend system like a CRM, Database or an API. These ‘intents’ are identified by utilizing Natural Language Processing (NLP) and the Machine Learning (ML). It turns out that some of these use cases can be enriched by allowing a user to upload an image. In such cases, you would want the conversation experience to take an action based on what exactly is in that image.

Let’s Make a Deal

It’s Here! How AI Robot Will Revolutionize Manufacturing

Though manufacturing has long been considered the industry with the highest degree of automation, fully-automated factories still seemed far away. However, AI-defined robotics is positioned to change that. How will robots with better dexterity and autonomous learning capabilities transform manufacturing processes and industry landscape? How should companies respond to the disruptive innovations that Robotics 2.0 brings?

Clean Machine Learning Code

Practical Software Engineering Principles for ML Craftsmanship. Machine learning (ML) pipelines are software pipelines after all. They are full of needless complexity and repetition. This is mixed with thick opacity, rigidity, and viscosity of design. With these issues, ML failures are growing in importance at an unprecedented pace. We have seen the self-driving cars hitting pedestrians in Arizona. We learnt about the gender bias of large scale translation system. We saw how simple masks hacked face id systems in smartphones. We heard about other ‘smart’ systems making bad decisions (e.g. Knight Capital). It is time to talk more about our responsibility in machine learning craftsmanship.