Original Source Here Fast.ai sejatinya adalah sebuah kelompok riset non-profit yang didirikan pada 2016 oleh dua orang Data Scientist bernama Jeremy Howard dan Rachel Thomas, dimana fokus terhadap pengembangan deep learning dan artificial intelligence. Mereka melakukan ini dengan menyediakan program kursus online terbuka (massive open online course (MOOC) yang diberi nama Practical Deep Learning for Coders atau Deep Learning Praktis untuk Pembuat Kode”. Kursus digelar secara gratis dan tidak memiliki prasyarat lain kecuali pengetahuan tentang bahasa pemrograman Python. Fast.ai memiliki 4 komponen utama yaitu: 1) kursus gratis, 2) library software, 3) riset mutakhir, 4) komunitas. AI/ML Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot via WordPress https://ramseyelbasheer.wordpress.com/2021/01/31/fast-ai-deep-learning/

Original Source Here Question classification — on cAInvas Finding the category of the question asked, i.e., the type of answer to be given. Photo by Mike Mirandi on Dribbble To answer a question, we need to understand the question and also the type of answer required. Different questions require different formats of answers. Categorizing questions based on answer formats help in addressing the questions better. For example, a question that starts with ‘how’ requires an answer that describes an event, a process, or procedure. This is called a descriptive answer. Questions with ‘what’ requires an answer that contains entities that fit the characteristics/definition given in the question. For a conversational A I model, identifying the answer type is as important to answering the question as is understanding the context of the question. Here, we categorize the questions into 6 categories — description, entity, abbreviation, numeric, human, and location. Implementatio...

https://cdn-images-1.medium.com/max/1557/0*9KqZX7G1EC2Rx8mk Original Source Here Review: Dual Attention Network for Scene Segmentation The aim of this article is to provide a brief overview of this paper Dual Attention Network for Scene Segmentation. Paper: The paper can be found online here . Publication: CVPR 2019 Institution: National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences Datasets: The paper uses Cityscapes Dataset , Pascal VOC Dataset , Pascal Context Dataset and COCO-Stuff Dataset . Overview: Fig.1. Dual Attention Network The paper asserts that although encoder- d ecoder architecture is a standard method for semantic segmentation and has achieved a lot of traction in recent years, it heavily relies on local information which may bring some bias as global information is not seen. The paper addresses this problem by capturing rich contextual dependencies based on the self-attention mechanism. The paper proposes D...

https://cdn-images-1.medium.com/max/2600/0*VxuNEHPDKJXjhY6L Original Source Here Layer 3 : Merchant Names Cleaning with spaCy By completing first two layers, we are able to solve some of the merchant names cleaning problems such as names of mis-spelling, different cases, missing characters/spaces and even some of the non-messy merchant inputs by simply returning a similarity score table. However, we are actually still in the phase of working with a rule-based cleaning engine, which means so far we still haven’t learned from the data. Furthermore, even by use of a typical machine learning model, the training phase still requires a large amount of time to perform feature engineering in creating more informative features. … potentially informative transaction-level features such as dollar amount and category, while also generating word-level natural language features such as word position within the label (e.g., 1st, 2nd), word length, proportion of vowels, consonants, and alph...

Original Source Here Review — t-SNE: Visualizing Data using t-SNE (Data Visualization) Visualizing High-Dimensional Data in Low-Dimensional Space t-SNE on MNIST Data set (From Google TechTalk by the First Author, Laurens van der Maaten, https://www.youtube.com/watch?v=RJVL80Gg3lA ) In this story, Visualizing Data using t-SNE , t-SNE , by Tilburg University, and University of Toronto, is briefly reviewed. It is a very famous paper by Prof. Hinton. In this paper: t-SNE is proposed, compared to SNE , it is much easier to optimize . t-SNE reduces the crowding problem , compared to SNE . t-SNE has been used in various fields for data visualization . This is a paper in 2008 JMLR with over 17000 citations . ( Sik-Ho Tsang @ Medium) It was also presented in 2013 Google TechTalk by author. AI/ML Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot via WordPress https://ramseyelbasheer.wordpress.com/2021/01/31/...

Original Source Here How to Speed up Your K-Means Clustering by up to 10x Over Scikit-Learn Chire , CC BY-SA 4.0 , via Wikimedia Commons K-Means Clustering is one of the most well-known and commonly used clustering algorithms in Machine Learning. Specifically, it is an unsupervised Machine Learning algorithm, meaning that it is trained without the need for ground-truth labels. Indeed, all you have to do to use it is set the number of desired clusters K , initialize the K centroids, and then the algorithm can be executed to get the classes. The beauty of K-Means lies in its simplicity: all it really does is compute the distances between points and group centers, resulting in a linear complexity O ( n ). This works perfectly fine with most datasets where you aren’t processing millions of data points. But that’s where we run into a problem: K-Means is slow when it comes to bigger datasets as there are just so many data points to compare. What’s worse is that the most popul...

Original Source Here Hypothesis Testing Pipeline So far, we have talked about the first two steps of hypothesis testing: setting up the null and alternative identify error types and set a significance threshold Now, we will look at a simple scenario using Python code. Below, we have the tips dataset from Seaborn which contains 244 records of clients coming to a restaurant. The dataset records bill and tip amount, table size, and other details. For simplicity, imagine you are the owner of the restaurant and the dataset holds the information for a single workday: All libraries are imported with their standard aliases. You calculate the average income for this day by dividing the results into two groups, dinner and lunch : Looks like on average, dinner-time clients paid more. Now, you wonder if this is just a random event specific to this day, or does this mean all future clients pay more for dinner? Let’s check this using a hypothesis test. Since we want to prove tha...