Explore every episode of the podcast Machine Learning Cafe
| Title | Pub. Date | Duration | |
|---|---|---|---|
| What is beyond PoCs? ML project-hurdles you should be prepared to take with Balázs Kégl - 016 | 14 Dec 2020 | 00:56:20 | |
In this episode, our guest was Balázs Kégl who is head of AI Research at Huawei Paris. We were talking about machine learning projects from the organizational point of view. We talked about the relationship between technical and non-technical people, and why are there so many POC projects why only a few of them is productionised? So, if you are a data scientist and you cannot convince your manager or your client to create a project, you should definitely listen to this episode. Moreover, we talked about some technical stuff at the end. Connect Balazs at Linkedin:
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info---
| |||
| Alternative Learning Methods 01 with Dmitry Krotov - 015 | 09 Sep 2020 | 00:57:17 | |
We continue with a micro-series, where we talk about alternative learning methods. In the first episode of this micro-series our guest is Dmitry Krotov, Researcher at MIT-IBM Watson AI Lab and IBM Research in Cambridge. Among many things, we will talk about an unusual learning rule, which has a degree of biological plausibility.
Paper: Unsupervised Learning by Competing Hidden Units Paper: Local Unsupervised Learning for Image Analysis: ---General Info--- About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| Finding the label errors with Cleanlab with Curtis Northcutt - 006 | 23 Jan 2020 | 00:55:44 | |
In this episode, I talked with Curtis Northcutt about his application cleanlab, with which you can find label errors in your dataset. Cleanlab computes cross-validated probabilities, the confident joint, and the statistics used in uncertainty estimation for dataset labels, and it ranks and sorts the labels by the probabilities of error, so you can easily find them in your dataset. Curtis' website: Curtis on LinkedIn: Cleanlab on GitHub: Cleanlab's blog: White Papers:
---General Info--- Podcast's Website: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| A New Deep Learning Optimizer called Ranger with Less Wright - 005 | 07 Jan 2020 | 00:36:54 | |
In this episode, we interview Less Wright, who has set 13 records on the fast.ai leaderboard, and we talk about one of his tricks is using cutting-edge optimizers, he also developed one, which is called Ranger. He talks also about different strategies of using Deep Learning optimizers, it is worth to take those into consideration. We also tested Ranger and in our case, it also outperformed Adam/RAdam variants. Co-host is Levente Szabados. Less' LinkedIn: https://www.linkedin.com/in/less-wright-22b59017/ Ranger's article: https://medium.com/@lessw/how-we-beat-the-fastai-leaderboard-score-by-19-77-a-cbb2338fab5c Less' medium articles: https://medium.com/@lessw His Github: https://github.com/lessw2020 Co-Host: Levente Szabados ---General Info--- Podcast's Website: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| Cutting-Edge Gradient Descent Variants with Levente Szabados - 004 | 19 Dec 2019 | 00:36:40 | |
In this episode, we are focusing on Deep Learning Optimizers, the different Gradient Descent Variants from vanilla GD to RAdam and Ranger. Levente tells us the story of GD from the simplest ones to the newest ones. This is part 1. Levente's Linkedin URL: An overview of gradient descent optimization Lookahead optimizer algorithm: Ranger optimizer by Less Wright: ---Copyright Info---
| |||
| Tree-based methods for procurement & are expert systems alive? with Renee Ahel - 003 | 12 Dec 2019 | 00:38:08 | |
In this episode, we interview Renee Ahel, who is a Lead AI Expert at Cirtuo and a freelance data scientist in Croatia. He is dealing with Machine Learning since 2003, and recently he is working on Tree-based methods, with which he solves procurement problems. He is also dealing with Expert systems, you will hear why. AI applications in procurement: Expert systems Kraljic matrix Bayes network powered PCOS (polycystic ovary syndrome) diagnostics - implemented by Croatian company Vingd for the Berlin startup Clue: Renee Ahel on LinkedIn: Data Science Croatia: | |||
| Can we predict the accuracy of a Neural Network? Yes, with the WeightWatcher tool by Charles Martin, Ph.D. - 002 | 07 Dec 2019 | 00:56:51 | |
In this Episode we talked about the deep neural networks and the spectral density of each layer's weights. It turns out, you can predict the accuracy ( and many more) with the WeigthWatcher application. We talk about the 5+1 Phases of learning and Heavy Tailed Self Regularization. Charles Martin, PhD on LinkedIn: During the episode we talked about these Why Deep Learning works? post by Charles Martin, 2015 Several Papers written by Charles Martin and Michael Mahoney: Newest blog post about weightwatcher, 2019 December: WeightWatcher on GitHub: How to reach out: To access their slack channel please contact Charles first.
| |||
| Progressively Growing GANs & Style GANs with Alexandr Honchar - 001 | 07 Dec 2019 | 00:46:56 | |
In this episode, we talked about the idea behind GANs in general and two special types of GANs: Progressively Growing GANs and Style GANs. LinkedIn URL of Alexandr Honchar: Companies where Alex works: Facebook: Blog about ECG pipeline: Progressively Growing GANs: StyleGAN:
---Copyright Info---
| |||
| "Monkeys with laptops" - the (hi)story of applying AI in real life with Denis Rothman - 014 | 23 Jun 2020 | 01:14:29 | |
In this "extra" episode, our guest is Denis Rothman, who is an Artificial Intelligence Specialist. He has spent 40 years in this industry, he is the author of the book called "Artificial Intelligence by Example". He is a teacher, a speaker, and he is an exceptional expert. LinkedIn: https://www.linkedin.com/in/denis-rothman-0b034043/ ---General Info--- About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| Conversational AI at Rasa with Vladimir Vlasov - 013 | 10 Jun 2020 | 01:03:24 | |
In this episode, our guest was Vladimir Vlasov, who is a senior machine learning researcher at Rasa. If you don't know Rasa, it's a company that is building a standard infrastructure for conversational AI. Vladimir is working on Rasa's machine learning-based dialogue tools which allow developers to automate contextual conversations. Vladimir on LinkedIn: https://www.linkedin.com/in/vladimir-vv-vlasov/ Rasa: https://rasa.com/ Conference on 18th June: https://www.l3-ai.dev YouTube channel: https://www.youtube.com/channel/UCJ0V6493mLvqdiVwOKWBODQ Algorithm Whiteboard: https://www.youtube.com/watch?v=wWNMST6t1TA&list=PL75e0qA87dlG-za8eLI6t0_Pbxafk-cxb Using Neural Networks to Find Answers in Tables:
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| What's real and what's hype? - Decades of ML with Eugene Dubossarsky - 012 | 21 May 2020 | 00:52:38 | |
In this episode, our guest is Eugene Dubossarsky, who is the chief data scientist at AlphaZetta and co-founder at multiple data science companies in Australia. He is a managing partner of the Global Training Academy and he is teaching too. Eugene is dealing with machine learning since the 80s, so you can imagine he has a very strong opinion about different topics in this industry. We talked about random forests, neural nets, boosting strategies, the importance of understanding data and statistics. We couldn't skip talking about the effects of the current and upcoming crisis either.
AlphaZetta: https://alphazetta.ai/ Reask Track cyclons: https://reask.earth/ The Prediction Machines book: Hopfield network: https://en.wikipedia.org/wiki/Hopfield_network Bengio talk: https://slideslive.com/38922304/from-system-1-deep-learning-to-system-2-deep-learning Finnian Lattimore and Cheng Soon Ong Paper
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| Extra: Machine Learning at John Deere with Marco Lemessi - 011 | 12 May 2020 | 00:47:29 | |
We recorded an extra and special (business related) episode, our guest is Marco Lemessi, who is a Machine Learning leader at John Deere. He will share some of the business aspects and also problems of the machine learning projects they are facing at one of the biggest agricultural companies in the world. He spent 18 years at John Deere, so you can imagine that he has a very broad perspective about business cases and data science and optimization problems at John Deere. He will talk about precision agriculture, problems of labeling on large dataset, legal aspects of artificial intelligence. Contact Marco on LinkedIn: https://www.linkedin.com/in/marcolemessi/ Marco's email address: LemessiMarco@johndeere.com Website of John Deere: https://www.deere.com/
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| The World’s first Quadruple Kaggle Grandmaster - Abhishek Thakur - 010 | 06 May 2020 | 00:44:39 | |
In this episode our guest is Abhishek Thakur, who is the Chief Data Scientist at Boost.ai in Norway. Abhishek has become the World’s first Quadruple Grandmaster on Kaggle. So we asked him about his experiences of the 150 competitions he has taken part in. So, what are the tricks here? How can someone participate in so many competitions, rank high and have a work besides these? Although he has been so successful you will see that he is a very humble person. Find Abhishek on Linkedin: https://www.linkedin.com/in/abhi1thakur/ Kaggle: https://www.kaggle.com/ Abhishek's user on Kaggle: https://www.kaggle.com/abhishek Subscribe him on YouTube: https://www.youtube.com/AbhishekThakurAbhi Website of Boost AI: https://www.boost.ai/
---General Info--- About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||
| Everything is just a wicked graph with Benedek Rozemberczki - 009 | 22 Apr 2020 | 00:50:12 | |
In this episode, our guest is Benedek Rozemberczki who is a Data Science Phd candidate at University of Edinburgh and also the owner of the github repo Karate Club where he implemented more than 30 scientific papers about Graphs and ML.
https://github.com/benedekrozemberczki/karateclub Papers on Graph Classification: https://github.com/benedekrozemberczki/awesome-graph-classification Papers on Node (Network) Embedding: https://github.com/chihming/awesome-network-embedding Papers on Graph Neural Networks: https://github.com/naganandy/graph-based-deep-learning-literature Papers on Graph Clustering: https://github.com/benedekrozemberczki/awesome-community-detection Deep Learning on Graphs: https://pytorch-geometric.readthedocs.io/en/latest/ Graph Kernels: https://github.com/jajupmochi/graphkit-learn Datasets https://snap.stanford.edu/
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/
---Copyright Info--- | |||
| Ratings and Marketing Attribution at Foursquare with Max Sklar - 008 | 04 Mar 2020 | 00:56:46 | |
In this episode our guest is Max Sklar who works as Engineering and Innovation Labs advisor at Foursquare. We talked about ratings and the problems about deciding if a rating is positive or negative, and the problems about different languages. In the second part of the show we talked about marketing attribution and causality. Max’s Linkedin: https://www.linkedin.com/in/max-sklar-b638464/ The Local Max Radio (podcast): https://www.localmaxradio.com/ Publication Timely Tip Selection for Foursquare Recommendations: http://ceur-ws.org/Vol-1247/recsys14_poster18.pdf Venue rating system blog post: https://medium.com/foursquare-direct/finding-the-perfect-10-how-we-developed-the-foursquare-venue-rating-system-c76b08f7b9b3
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info---
| |||
| MISH Activation Function with Diganta Misra - 007 | 02 Feb 2020 | 00:50:29 | |
In this episode, my guest is Diganta Misra, who is the founder of Landskape AI, which is a research lab aimed at solving the most challenging questions of Deep Learning. Diganta is a Mathematician who invented the activation function called MISH, which beats Google's activation function called Swish in most computer vision tasks. He released his paper last summer and the project got already 600 stars on Github. He is already working on the next activation function called SharkFIN, which will be even better than MISH. Contact Diganta on LinkedIn: https://www.linkedin.com/in/misradiganta/ Landskape AI webpage: landskape.org MISH Paper: MISH project on Github:
About the co-host: Levente Szabados is a Deep tech leader, consultant, and manager with a special interest in artificial intelligence, cognitive sciences, data science, and deep learning. He is teaching various ML courses at the Frankfurt School of Finance and Management. About the podcast: The podcast was created to serve the technical community with details of ML algorithms, use-cases. The episodes are recorded in Budapest, Hungary, Europe. Website of the podcast: http://machinelearningcafe.org/ ---Copyright Info--- | |||