Learning to generate videos with uncertain futures

TL;DR: This post provides a high-level overview of the video generation model described in Stochastic Video Generation with a Learned Prior, which is capable of generating video sequences with multiple futures. Video Generation as Self-Supervised Learning task Supervised deep learning models have proven to yield groundbreaking results in the recent past on hard tasks like real-time … Read more Learning to generate videos with uncertain futures

Mapping Physical Activity with R, Selenium and Leaflet

We all know that exercise is one of the most important factors in our mental and physical health. And with the new year fast approaching, an emphatic declaration to Work Out More! is sure to top many resolution lists. But figuring out how to actually accomplish this can be difficult. While January is the most … Read more Mapping Physical Activity with R, Selenium and Leaflet

Lumiere London 2018 (Part 3): Computer Vision

Part 3: Analysing 5,000 Flickr images using computer vision Introduction In this final blog post of the series, I apply computer vision techniques to understand 5,000 Flickr images about the Lumiere London 2018, a huge light festival which took place in London earlier in January this year. During Lumiere London 2018, more than 50 public artworks … Read more Lumiere London 2018 (Part 3): Computer Vision

Explainable AI vs Explaining AI — Part 1

Despite the recent remarkable results of deep learning (DL), there is always a risk that it produces delusional and unrealistic results due to several reasons such as under-fitting, over-fitting or incomplete training data. For example the famous Move 78 of the professional Go player Lee Sedol which caused a delusional behavioral of Alpha Go, adversarial … Read more Explainable AI vs Explaining AI — Part 1

Understanding AI and ML for Mobile app development

Last time I published this blog where I explained about one application of AI and ML — ‘Vision’ and also explained briefly about using ML kit in mobile development which is a cloud platform offered by Google to integrate ML features in Android and iOS apps. This article is prequel to that one and in this I … Read more Understanding AI and ML for Mobile app development

SD-WAN Link Switch as Reinforcement Learning experiment with Deep Q-Learning

Credit — https://usblogs.pwc.com/emerging-technology/deep-learning-ai/ ‘Deep Q’ or Deep Q-Learning is a well-known algorithm in reinforcement learning which approximates Q Value of an MDP system with deep neural network. In this article I have explored the same algorithm in solving the link switch problem in SD-WAN network for which I already have developed an AI-gym based on Mininet (see … Read more SD-WAN Link Switch as Reinforcement Learning experiment with Deep Q-Learning

All birds are black

A simple way to think about bias-variance trade-off Photo by Hannes Wolf on Unsplash I’ve come across multiple approaches and philosophies for building models to represent real world relationships. My statistics professor was relentless in emphasising Occam’s Razor and parsimony. Social scientists are obsessed with finding causal relationships in models, often through experiments. Don’t go there, Simba! … Read more All birds are black

Forecasting with Prophet

How to make high quality forecasts The origin of Prophet When we think of forecasting we often think of weather forecasts, but it is also used by many organizations in supply chain management, sales and economics. Forecasts are used to guide policymakers and play an important role in shaping business decisions (e.g. Federal Reserve adjusting interest … Read more Forecasting with Prophet

30 Data Science Punchlines

A holiday reading list condensed into 30 quotes For those who like brainfood on your vacation, here’s a handy index of all my articles from 2018 boiled down to 30 (occasionally cheeky) punchlines to help you avoid/cause awkward silences at family events and holiday parties. Sections: Data Science and Analytics, ML/AI Concepts, How Not To Fail … Read more 30 Data Science Punchlines

Why Machine Learning is the BEST field in the world

A few years ago, when I was a junior software engineer, I worked on a problem with one of our algorithm developers. I thought that I found the breaking point: there was an algorithm that did something wrong. I asked the developer why the algorithm did what it did, and the answer I got was: … Read more Why Machine Learning is the BEST field in the world

The Mathematics Behind Principal Component Analysis

Introduction The central idea of principal component analysis (PCA) is to reduce the dimensionality of a data set consisting of a large number of interrelated variables while retaining as much as possible of the variation present in the data set. This is achieved by transforming to a new set of variables, the principal components (PCs), … Read more The Mathematics Behind Principal Component Analysis

Investigating the variance of the world chess championship final

A statistical analysis on Carlsen’s final match approach 2018 World Chess Championship logo showing 5 overlapping arms above chessboard holding or moving chess pieces. While the FIDE World Chess Championship was going on, I found the following problem: Say one of the players is better than his opponent to the degree that he wins 20 percent of … Read more Investigating the variance of the world chess championship final

Which hypothesis test to perform?

Overview of the various hypothesis tests with an example of a one-sample t-test The objective of statistics is to make inferences about a population based on information contained in a sample. The numerical measures used to characterize populations are called parameters. The population parameters are : μ: mean M: median σ: standard deviation π: Proportion Most … Read more Which hypothesis test to perform?

20 Years of Data, 10 Conclusions

I got my first “real” job in 1999, working for a power company in Copenhagen (shout out to Lars) creating electricity pricing reports in Excel. Since then I’ve worked for small companies, start-ups and large companies across a range of industries. I’ve worked with passionate founders as well as hired guns and I’ve sat at … Read more 20 Years of Data, 10 Conclusions

Myth-busting about Data Science with Simon Greiner

Cesar Viteri on www.unsplash.com In her initial blog, Anita Lakhotia asked the question: What does a Data Scientist do all day? This is the most frequent question she gets asked by non-data-scientists. Thinking about data scientists she has personally come across during meet-ups, hackathons or blogs, it is difficult to give just one answer. Therefore, … Read more Myth-busting about Data Science with Simon Greiner

You can’t just Google everything

And other things I wish I knew before I started my latest Data Science project Photo by Chris Ried on Unsplash I got started on a new data science project a few weeks ago and expectedly, it’s been h̶o̶r̶r̶i̶f̶i̶c̶ absolutely enlightening. I’ve made a few tentative explorations into data science and machine learning projects before. But I … Read more You can’t just Google everything

Multi-Class Classification in Text using R

This blog is in continuation to my NLP blog series. In the previous blogs, I discussed data pre-processing steps in R and recognizing emotions present in ted talks. In this blog, I am going to predict the ratings of the ted talks given by viewers. This would require a multi-class classification and quite a bit … Read more Multi-Class Classification in Text using R

AI for Business

This is going to be a short, to-the-point article. First I’ll talk about the problems with AI right now, then the problems with understanding and applying AI in business scenarios, finally a first glance on the solutions I’m thinking. Part I. The AI world I hate this image 🙂 Since the beginning of humanity as a society, … Read more AI for Business

Shrinkage Estimators: Shrinking statistical estimates

In this article series on how to optimize portfolios, we have looked at the existence of market invariants, estimating distribution of returns using nonparametric and maximum likelihood. Now we discuss a method of estimating the probability distribution using shrinkage estimators. For those interested in optimizing portfolios, look at OptimalPortfolio. I must agree, the name shrinkage … Read more Shrinkage Estimators: Shrinking statistical estimates

Implementation of Uni-Variate Linear Regression in Python using Gradient Descent Optimization from…

Learn, Code and Tune…. Regression is an example of Continuous Classification of Data or data-points in feature-space. Francis Galton invented the usage of Regression Line in 1886 [1]. As the name suggests, “Linear”, this means that the hypothesis regarding the Machine Learning Algorithm is linear in nature or simply a linear equation. Yeah!! it’s a … Read more Implementation of Uni-Variate Linear Regression in Python using Gradient Descent Optimization from…

Review: DeepMask (Instance Segmentation)

1. Model Architecture Model Architecture (Top), Positive Samples (Green, Left Bottom), Negative Samples (Red, Right Bottom) Left Bottom: Positive Samples A label yk=1 is given for k-th positive sample. To be a positive sample, two criteria need to be satisfied: The patch contains an object roughly centered in the input patch. The object is fully contained in … Read more Review: DeepMask (Instance Segmentation)

A different kind of (deep) learning: part 2

Self Supervised learning: generative approaches Intro In the previous post, we’ve discussed some self supervised learning articles, along with some attempts to strive towards the “holy grail”: exploiting the almost unlimited number of un-annotated images available wherever to generalize for other tasks. And hopefully, get closer to the currently unmet benchmark of ImageNet pre-training. Surprisingly, … Read more A different kind of (deep) learning: part 2

Logistic Regression For Facial Recognition

Facial recognition algorithms have always fascinated me, and wanting to flex my newfound logistic regression skills on some data, I created a model based on a dataset I found called “Skin Segmentation.” As noted in its description, the data in Skin Segmentation were collected “by randomly sampling B,G,R values from face images of various age … Read more Logistic Regression For Facial Recognition

Planet Beehive

Exploring our Activities Now that we have created our Scoring Measure, we can start exploring the data. Let’s start by comparing global regions’ activity rating with the average number of reviews per activity: Fig. 8 Scatter plot generated with Seaborn in Python Now this is getting interesting! Some very particular facts here that might just go … Read more Planet Beehive

Machine Learning and Music Classification: A Content-Based Filtering Approach

Using the Librosa Python Library, KNN, and Random Forest to Classify Music In my previous blog post, Introduction to Music Recommendation and Machine Learning, I discussed the two methods for music recommender systems, Content-Based Filtering and Collaborative Filtering. The collaborative filtering approach involved recommending music based on user listening history, while the content-based approach used an … Read more Machine Learning and Music Classification: A Content-Based Filtering Approach

Get Smarter with Data Science — Tackling Real Enterprise Challenges

Introduction The ‘Data Science Strategic Guide — Get Smarter with Data Science’ is envisioned as a series of articles, which serve to be more of a strategic guide depicting essential challenges, pitfalls and principles to keep in mind when implementing and executing data science projects in the real-world. We also focus on how you can get maximum … Read more Get Smarter with Data Science — Tackling Real Enterprise Challenges

Measuring pedestrian accessibility

Walkable neighborhoods are great for health, happiness and economic growth. Cities around the world that want to draw a talented young workforce increasingly focused on creating a good pedestrian experience. How could we measure and map walkability using data science tools? This blog suggests an approach drawing on Pandana, an excellent Python library developed by … Read more Measuring pedestrian accessibility

Classifying Skin Lesions with Convolutional Neural Networks

Imagine this. You wake up and find a frightening mark on your skin so you go to the doctor’s office to get it checked up. They say it’s fine so you go home and don’t worry about it for a couple months, but then you have a throbbing pain from that spot — it looks ugly and … Read more Classifying Skin Lesions with Convolutional Neural Networks

Building sentence embeddings via quick thoughts

Introduction to Quick Thoughts In previous story, I shared skip-thoughts to compute a sentence embeddings. Today, we have another unsupervised learning approach to compute sentence embeddings which is Quick Thoughts. Logeswaran et al. (2018) introduced quick-thoughts approach to retrieve sentence embeddings for downstream application. After reading this article, you will understand: Quick-Thoughts Design Evaluation Experiments Reference … Read more Building sentence embeddings via quick thoughts

Quality inspection in manufacturing using deep learning based computer vision

Improving yield by removing bad quality material with image recognition Author: Partha Deka and Rohit Mittal Automation in Industrial manufacturing: Today’s increased level of automation in manufacturing also demands automation of material quality inspection with little human intervention. The trend is to reach human level accuracy or more in quality inspection with automation. To stay … Read more Quality inspection in manufacturing using deep learning based computer vision

Do you need a graduate degree for data science?

Maybe so. Maybe not. I’ll level with you: I’m a PhD dropout. I’ve gotten a lot of mileage out of that title, by the way: it hints that I’ve done a lot of grad school, but still maintains the aura of badassery that only the word “dropout” can provide. In some ways, it’s the ultimate humble … Read more Do you need a graduate degree for data science?

Applying Logistic Regression to PubMed

In my last article, we looked at how to get meaningful insights from a huge collection of medical articles gathered from PubMed, a free archive of biomedical and life sciences literature. This time, we are going to continue working with medical articles to create something completely different. We will use Supervised Machine Learning to identify … Read more Applying Logistic Regression to PubMed

Decoding Gene Set Variation Analysis

Characterising biological pathways from gene expression data Gene Set Variation analysis is a technique for characterising pathways or signature summaries from a gene expression dataset. GSVA builds on top of Gene Set Enrichment analysis where a set of genes is characterised between two condition groups defined in the sample. GSEA (Gene set enrichment analysis) works … Read more Decoding Gene Set Variation Analysis

I Analyzed 122 Hours of Holiday Radio

How many times will you be forced to hear “Wonderful Christmastime”? 122 hours, 1,510 tracks. Only 80 original songs. Source: 106.7 LiteFM; 11/30/2018–12/5/2018; Download the data. It starts well before Thanksgiving. This year it was November 16th at 5pm to be precise. That’s when New York’s WLTW 106.7 LiteFM makes a hard switch to an all-Christmas … Read more I Analyzed 122 Hours of Holiday Radio

Serverless Distributed Data Pre-processing using Dask, Amazon ECS and Python (Part 1)

Source: unsplash.com Dec 18, 2018 The quality and accuracy of machine learning models depend on many factors. One of the most critical factors is pre-processing the dataset before feeding it into the machine learning algorithm that learns from the data. Therefore, it is critical that you feed them the right data for the problem you … Read more Serverless Distributed Data Pre-processing using Dask, Amazon ECS and Python (Part 1)

Think Machine Learning and AI Won’t Impact Your Networking Product — Think Again!

Using ML and AI as a force-multiplier will be a significant competitive advantage for networking product teams Photo by Hitesh Choudhary on Unsplash Machine learning and related techniques have seen tremendous advances in the last few years. And while at times it might feel that there’s a lot of hype surrounding the space, it’s clear that … Read more Think Machine Learning and AI Won’t Impact Your Networking Product — Think Again!

Friend Recommendation Using Heterogeneous Network Embeddings

Imagine Snoopy without Woodstock or Calvin without Hobbes, Friends without Rachel, Batman without Robin or Mowgli without Baloo. Social platforms thrive on the ability of the members to find relevant friends to interact with. The network effect is what drives growth or time spent and daily active users on the application. This is even more … Read more Friend Recommendation Using Heterogeneous Network Embeddings

Multimodal Deep Learning

Fusion of multiple modalities using Deep Learning Being highly enthusiastic about research in deep learning I was always searching for unexplored areas in the field (Though it is tough to find one). I had previously worked on Maths word problem solving and many such related topics. The challenge of using Deep Neural Networks as black boxes … Read more Multimodal Deep Learning

Docker Without the Hassle

How to Use: Slightly Friendlier Version First install docker. Instructions for your machine can be found here. The docker getting started guide is useful for learning how docker works, although we don’t need the details to use it effectively with repo2docker Make sure docker is running. If docker run hello-world shows the message Hello from … Read more Docker Without the Hassle

Distributed TensorFlow using Horovod

Reduce training time for deep neural networks by using many GPUs Marenostrum Supercomputer — Barcelona Supercomputing Center https://bsc.es (This post will be used in my master course SA-MIRI at UPC Barcelona Tech with the support of Barcelona Supercomputing Center) “Methods that scale with computation are the future of Artificial Intelligence” — Rich Sutton, father of reinforcement learning (video 4:49) In … Read more Distributed TensorFlow using Horovod

A journey into supervised machine learning

Some practical examples, tips, and thoughts on supervised ML Earlier this year, through my MBA program at Cornell Tech, I took a great intro course on Machine Learning with a fantastic professor, Lutz Finger. Lutz’s course inspired me to dig even deeper into ML and AI, so I recently started a hands-on Introduction to Machine … Read more A journey into supervised machine learning