Recently I’ve been working with manufacturing customers (both OEM and CM) who want to jump on the bandwagon of machine learning. One common use case is to better detect products (or Device Under Test/DUT) that are defective in their production line. Using machine learning’s terminology, this falls under the problem of binary classification as a DUT can only **pass** or **fail**.

However, training a binomial classifier that requires samples from both cases of **pass **and **fail **turns out to be impractical. There are two reasons why.

- Imbalanced data. Manufacturing processes are optimized to produce devices with high yield i.e. as…

During my literature review, I stumbled upon an information-theoretic framework to analyse the link prediction problem (Tan et al. (2014), Kumar and Sharma (2020)). For an overview of what link prediction is, read my previous article here. The basic idea is to predict unseen edges in a graph. Such edges could represent citing relationship, friendship, or dependencies in different networks. The framework models such task as an information theory problem, where an edge is more likely to form over a pair of nodes if it has a high probability (low uncertainty/low information) based on the chosen model.

In this article…

Answer: They can all be mathematically formulated as a graph link prediction problem!

In short, given a graph G (V, E) with |V| vertices and |E| edges, our task is to predict the existence of a previously unknown edge e_12 **∉** E between vertices v_1, v_2 **∈ **V. We can then use the link prediction model to, for instance, recommend the two vertices to each other. …

I’ve recently been looking for an open-source, distributed graph database, as I need to store a large graph data somewhere persistently. My main requirement is that I’d like to have as much control as possible over the underlying storage and indexing system behind such aforementioned database.

I stumbled upon JanusGraph, a graph database project under the Linux Foundation, that is built on top of Apache TinkerPop, including the Gremlin query language. Tinkerpop powers a lot of the other graph databases out there too, such as neo4j, Amazon Neptune, DataStax, Azure Cosmos DB etc… I chose JanusGraph because of its plug-and-play…

Pada bulan Juni 2020 yang lalu, kami diundang oleh International Telecommunication Union untuk memberikan presentasi tentang peran Alva Energi di Indonesia dalam sebuah virtual conference tahunan yang bernama World Summit on the Information Society.

Dalam webinar tersebut, kami bercerita lebih tentang perkembangan energi terbarukan di Indonesia dan juga peran aktif yang telah kami lakukan selama ini. Rekaman webinar tersebut dapat diakses di website resmi WSIS berikut ini: https://www.itu.int/net4/wsis/forum/2020/Agenda/Session/194.

Find the latex-written version of this article here

The following document provides a whirlwind tour of some fundamental concepts in geometric deep learning. The mathematical derivations might not be as rigorously shown and some equations are stated without proofs. This is done intentionally to keep the document short yet comprehensive enough. For deeper analysis, please refer to Bronstein et al. (2017), Gong (2018), Kipf et al. (2017), Hammond et al. (2011), Shuman et al. (2011) and http://geometricdeeplearning.com/. **This is a living document, so please let me know if you find any errors or inconsistency, I will fix them as soon…**

This year marks the 17th anniversary of CUTEC. So I thought I’ll reach out to our former committees to see how they are doing and perhaps ask for a few thoughts on how the CUTEC experience, in hindsight, contributes to their life. The responses are surprisingly very encouraging! Do take a look at some testimonies from our alumni below. If you’re one of them, we’d just like to say a huge thanks for paving the way for us future committees! If your name isn’t here yet, do reach out to us as we’re keen to hear from you!

I recently learned about Gaussian Process (GP) and how it can be used for regression. However, I have to admit that I had a hard time grasping the concept. It was only after I derived the equations and tried going through a few samples did I manage to start deciphering what this whole idea is about.

a Gaussian process is a collection of random variables, any finite number of which have (consistent) Gaussian distributions. (Rasmussen)

Let’s start by looking at a simple problem. Suppose we observe 2 pairs of data points (x1, y1) and (x2, y2). Now given a new…

Servitization is a phenomena where manufacturing firms shift from selling pure products to offering solutions (services) instead. Neely (2013) provides a brief introduction on how companies across industries are adopting this business model.

What’s interesting is that servitization also results in a less clear boundary between manufacturing and service firms (read Vandermerwe and Rada, 1988 for a good review). In fact, we can see that some aspects of manufacturing and services can be abstracted into similar mathematical concepts. For example, compare an assembly line with a group of people lining up to buy food. …

This post is a summary of the theory of nonlinear dynamics and chaos that I have recently learned from an online course by the Santa Fe Institute. All materials are credited to the institute alone.

Dynamics is a branch of mathematics that studies how systems change over time. Up until the 18th century, people believed that the future could be perfectly predicted given that one knows* “all forces that set nature in motion, and all positions of all items of which nature is composed” *(that one being is referred to as a Laplace Demon).

Now, provided that we believe that…

Machine Learning | Network Science | Supply Chain and Manufacturing Analytics | eekosasih.com