What is a Good Metric?

Analytics is about tracking the metrics that are critical for your business. Usually this metrics can give a picture of your business and relate to your business model, things like where the money comes from, where is it going, how many customers are com

Mathematics Dataset

This dataset code generates mathematical question and answer pairs, from a range of question types at roughly school-level difficulty. This is designed to test the mathematical learning and algebraic reasoning skills of learning models.

Economic Value of Learning and Why Google Open Sourced TensorFlow

Blog key points:
• Google open-sourced TensorFlow to gain tens of thousands of more users across hundreds (thousands) of new use cases to improve the predictive effectiveness of the platform that runs Google’s business.
• In the digital economy, the economies of learning are more powerful than the economies of scale.
• Organizations must avoid building orphaned analytics – one-off analytics developed to address a specific business need but are never ‘operationalized’ or packaged for re-use across the organization.
• Instantaneous acceleration is the acceleration of an object at a specific moment in time as measured by the second derivative.

Neural Networks: why do they work so well? Part I

Neural networks are a fundamental model of deep learning. Their techniques underly most of the more complex models of the field. It’s therefore very important for a practitioner of deep learning to have an intuitive grasp of what neural networks do. Unfortunately, the way neural nets are sometimes taught hides why they work. Many explanations get caught up in ‘brain-inspired’ words (like layer, activation, neuron) or cool-looking network diagrams (like the one below). These might sound and look nice, but they aren’t really that useful. In this article, we’re going to talk about what’s really going on with neural nets. In place of big words and cool diagrams, we’ll focus on the fairly basic math that makes the model work so well.

Weight Initialization in Neural Networks: A Journey From the Basics to Kaiming

I’d like to invite you to join me on an exploration through different approaches to initializing layer weights in neural networks. Step-by-step, through various short experiments and thought exercises, we’ll discover why adequate weight initialization is so important in training deep neural nets. Along the way we’ll cover various approaches that researchers have proposed over the years, and finally drill down on what works best for the contemporary network architectures that you’re most likely to be working with.

The Data Product Design Thinking Process

Curiosity is a major driver for all of us. We are constantly carrying out research into processes, causes and effects. We want to understand how companies, operational processes and economic relations work, what steps and parts they consist of, why things happen and how everything is interrelated. Once we have analyzed everything, we can use this knowledge to exert influence on the world in a positive way. Data is an increasingly important factor in the search for correlations. The right data (smart data instead of big data) shed more light on the darkness. Without data, an accurate, systematic analysis and precise design of the world around us is no longer conceivable. Intuition and gut feeling alone are no longer enough.

Real world implementation of Logistic Regression

Classification techniques are an essential part of machine learning and data mining applications. Approximately 70% of problems in Data Science are classification problems. A popular classification technique to predict binomial outcomes (y = 0 or 1) is called Logistic Regression. Logistic regression predicts categorical outcomes (binomial/multinomial values of y), whereas linear Regression is good for predicting continuous-valued outcomes (such as the weight of a person in kg, the amount of rainfall in cm). I have divided this article into 3 parts. In the first part we’ll take a look at some of the important concepts of Logistic Regression, In the second part we’ll build a Binary classifier and in the third part, we’ll build a Multiclass classifier.

NLP Keras model in browser with TensorFlow.js

How to use Your Keras Model in Browser With Tensorflow.js. In This Article I’ll Attempt to Cover Three Things:
1. How to Write Simple Named-Entity Recognition Model – typical Natural Language Processing (NLP) Task.
2. How to Export This Model to TensorFlow.js Format.
3. How to Make Simple WEB Application for Searching Named-Entity in String Without Back-End.

DevOps for machine learning research workflows

Doing machine learning research for the first time? Or perhaps you’ve done this at a large institution and now you’re trying to figure it out on your own? This article is for you! In recent years, I’ve been working on several projects that require focused and disciplined empirical research. Most of my work is done solo with varying degrees of indirect input from other people. Because of this, I’ve had the opportunity to set up my research workflows completely from scratch. Because a lot of research is spent running controlled experiments and quickly iterating on ideas, it’s crucial to have an organized and robust research environment. From the many hours I’ve spent (and many mistakes I’ve made), I’ve discovered that improving my research workflow involves asking myself these two questions: 1. What can I do to improve my confidence in the results of my work? 2. How do I shorten my feedback cycle so that I can iterate faster? Each of the three ideas presented below answers one or both of the questions above.

The Data Fabric for Machine Learning. Part 2: Building a Knowledge-Graph.

I’ve been talking about the data fabric in general, and giving some concepts of Machine Learning and Deep Learning in the data fabric. And also gave my definition of the data fabric: ‘The Data Fabric is the platform that supports all the data in the company. How it’s managed, described, combined and universally accessed. This platform is formed from an Enterprise Knowledge Graph to create an uniform and unified data environment.’ If you take a look at the definition, it says that the data fabric is formed from an Enterprise Knowledge Graph. So we better know how to create and manage it.

Familiarity With Coefficients Of Similarity

When you were doing a project on the recommendation system or the semantic segmentation of the images, you must have come across similarity scores. Based on these similarity scores, you predicted that this product is similar to that product or how much the predicted segmented image is similar to the ground truth. Similarity metrics are important because these are used by the number of data mining techniques for determining the similarity between the items or objects for different purposes as per the requirement such as, clustering, anomaly detection,automatic categorization, correlation analysis.This article will give you a brief idea about different similarity measures without going too much into the technical details. The main focus of this article is to introduce you to the below similarity metrics,
1. Simple matching coefficient (SMC)
2. Jaccard index
3. Euclidean distance
4. Cosine similarity
5. Centered or Adjusted Cosine index/ Pearson’s correlation
Let’s start!

A short tutorial on Fuzzy Time Series – Part III

Interval and probabilistic forecasting and non-stationary methods. Hi folks! It was a long time since I published the first and second parts of this tutorial. Meanwhile I had the opportunity to talk with many people that applied FTS methods on several distinct fields, and they helped me to improve pyFTS library by fixing some bugs and expanding its features and usability. We have been dealing with uncertainty since the beginning of this tutorial. We found ways to model and describe the temporal behavior of the time series using fuzzy sets, but we did not take into account the uncertainty of our forecasts. The forecasting uncertainty is like a fog: if you are looking closer you will see blurred and it may not affect you too much, but if you want to look more ahead… So, once again it is time to go deeper and explore more forecasting types and also new methods, this time I will focus on non-stationary time series, concept drifts, etc.