Albert Cheng

🚀 Exciting News: Open-Sourcing a Game-Changing Keras Data Preprocessing Tool! 🐍📊 After years of closed development, I'm thrilled to announce the release of my TensorFlow Keras-based data preprocessing package (KDP) v1.8.0 (finally had some time to release it). This tool streamlines ML workflows, making data preparation more efficient and scalable. Key features: • Built on TensorFlow Keras • Innovative use of Keras preprocessing layers • Suitable for both beginners and seasoned data scientists • Integrates directly as first layers of your Keras models or as a separate one ! You can read some more here: https://lnkd.in/gRY2AFGe I'm open-sourcing this project to empower the community. Your contributions can help take it to the next level! Check it out and let's revolutionize data preprocessing together! 💡 #DataScience #MachineLearning #OpenSource #Python #TensorFlow

35

I’m thrilled to collaborate with Josep from DataBites to share a comprehensive guide of top resources in ML. Josep Ferrer is a veteran technical writer of machine learning having published more than 50+ articles in Kdnuggets and towardsDataScience. Together, we've curated an essential list of resources to help you build a strong foundation in ML. The guide is divided into two parts. In Part 1, we share key theoretical resources: • Top-rated online courses • Must-read books. • Popular podcasts and tutorials to get you started. These high-quality resources are designed to equip you with the knowledge and confidence to tackle the challenges ahead in your ML journey. Checkout the article here: https://lnkd.in/gVDS_Uaw Stay tuned for Part 2 on Saturday, where we'll dive into practical, hands-on learning resources, some classic projects, and advanced techniques to help you apply your knowledge and build a standout portfolio. Don't miss out! Check out Part 1 now and take the first step. -- 𝘐𝘧 𝘺𝘰𝘶 𝘭𝘪𝘬𝘦𝘥 𝘵𝘩𝘦 𝘢𝘳𝘵𝘪𝘤𝘭𝘦, 𝘢𝘭𝘴𝘰 𝘤𝘰𝘯𝘴𝘪𝘥𝘦𝘳 𝘴𝘶𝘣𝘴𝘤𝘳𝘪𝘣𝘪𝘯𝘨, 𝑙𝑖𝑛𝑘 𝑏𝑒𝑙𝑜𝑤 𝑚𝑦 𝑝𝑟𝑜𝑓𝑖𝑙𝑒 𝑝𝑖𝑐𝑡𝑢𝑟𝑒 ☝️ #MachineLearning #LearningJourney #MLResources

69

20 Kommentare

Want to give a talk about Managing ML project at the largest ML conference in the world? Submit your talk proposal to the KDD Workshop on Applied ML Management. The workshop focuses on real use-cases and take-aways from managing machine learning projects, teams, and products. It is a great opportunity to engage in discussions with industry leaders. More info is here: https://lnkd.in/ewrxKniT See you at KDD-2024 in Barcelona! #mlmanagement #kdd #amazonscience #conferences #ml #ai #science

44

1 Kommentar

Unlock the full potential of your Large Language Models (LLMs) with Intel® Extension for PyTorch (IPEX) and the Intel® LLM Library for PyTorch (IPEX-LLM). Download this whitepaper to explore the optimization of LLMs, including their performance, resource utilization, and response times in real-world applications. Link - https://intel.ly/3BBJ4ey

1

Data Scientists, This is BIG! Snowflake's Arctic LLM is Here – and Open-Source! 🤯 Get your hands on it now: Interactive Demo: Explore Arctic's capabilities with this Streamlit app → https://lnkd.in/gJ7y38Ng Replicate: Ready to experiment? Find it on Replicate → https://lnkd.in/gyDXiMju Why This Matters: Arctic lets you quickly build demos of your models. This is a game-changer for showcasing your work and making your models more accessible! Huge thanks to Snowflake for making this happen! Are you excited to try Arctic? Let me know in the comments! 👇 #LLM #OpenSource #DataScience #Streamlit #Snowflake

8

Hello Network! My review of the month of September is finally out! AI have been through +4k papers and here is the selection that stood out after i've done my part :). Topics include: - Recommender Systems with Mechanism Design - Ranking Losses for Auctions with Welfare Maximisation, - What's the Motivations behind "Democratising" in Open-Source AI - Ethics on Brain Interfaces - Minimal Viable Ethics (yes there is a MVP equivalent for Ethics) https://lnkd.in/dMrH8qYb If you like the work please share or comment so as to the algorithm to pick it up :) #MachineLearning #AI #Ethics #Technology #StateOfTheArt #Innovation

18

Dynamic Decoding: A Practical Step Forward for LLMs The concept of Adaptive Decoding via Latent Preference Optimization offers a thoughtful way to improve language model outputs by dynamically selecting decoding temperatures. By building on frozen LLMs, the ADAPTIVEDECODER extracts rich contextual insights from hidden states, efficiently adapting to tasks like reasoning and storytelling. Its reliance on preference optimization (LPO) instead of manual tuning reduces complexity while improving flexibility. With minimal training data, the method achieves impressive adaptability across diverse tasks, marking a practical and scalable improvement for LLM applications. This approach feels like a natural progression toward more tailored and efficient language model usage. https://lnkd.in/e-_JRtAP

1

Gigabites is back! While LLMs are hot at the moment, this Meetup focuses on challenges in traditional Machine Learning. Don’t worry, these talks are not just for AI experts but for anyone with an interest in data (and amazing pizza)!

28

1 Kommentar

Excited to announce a new webinar at DataTalksClub about Inventory Optimization in E-commerce (inspired by our work at Zalando)! From modeling your cost/profit equation to optimizing it for millions of articles, there is a long way to go! Lot of scientific AND MLOps decisions to make! It's not only about finding optimal decisions for your inventory but doing it continuously and doing it at an incredible speed! The talk will cover: - ​The problem of inventory optimization in e-commerce. - What decisions do we need to optimize? (What and When to replenish?) ​- ​What types of inputs and forecasts do we need? -​ How do you model your inventory's cost equation? - How can you incorporate probabilistic forecasts in real-world inventory modeling using Monte Carlo simulations and gradient-free optimization? - ​How can you use Numba and cloud infrastructure to scale these techniques to millions of articles per day? Join us (Myself, and Andreas Syrén ) on Tuesday 2nd July at 4 pm! https://lu.ma/blpa8il4

31

1 Kommentar

Soda: Ideal Choice for Sharing Data Quality Results with Non-Technical Users Recently, I had the opportunity to explore various Data Quality tools, including Great Expectations, Deequ, dbt_expectations, and Soda. Each tool offers excellent test coverage, supports custom test cases, and integrates seamlessly within different environments. However, if you need to share the results of your test cases with non-technical users and visualize them, the cloud version of Soda stands out as the best option. Sharing test case results with non-technical users is crucial for several reasons: 1. 𝗙𝗮𝗰𝗶𝗹𝗶𝘁𝗮𝘁𝗲 𝗗𝗮𝘁𝗮 𝗙𝗶𝘅𝗶𝗻𝗴: Provides non-technical teams with access to failed records if they are responsible for fixing these issues. 2. 𝗜𝗺𝗽𝗿𝗼𝘃𝗲𝗱 𝗧𝗿𝗮𝗻𝘀𝗽𝗮𝗿𝗲𝗻𝗰𝘆: Helps stakeholders understand the quality of the data driving their decisions. 3. 𝗘𝗻𝗵𝗮𝗻𝗰𝗲𝗱 𝗖𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻: Facilitates better communication and collaboration between technical and non-technical teams. 4. 𝗜𝗻𝗳𝗼𝗿𝗺𝗲𝗱 𝗗𝗲𝗰𝗶𝘀𝗶𝗼𝗻-𝗠𝗮𝗸𝗶𝗻𝗴: Empowers all team members to make data-driven decisions with confidence. Soda's recent success was not a surprise for me: https://lnkd.in/dkpQ2Wqn Soda's official documentation: https://docs.soda.io/ #DataQuality #DataValidation #DataEngineering #GreatExpectations #Deequ #dbt_expectations #Soda #DataTransparency #DataVisualization #DataGovernance

5

FlexAttention: Simplifying Attention in PyTorch _________________________________________________________________ FlexAttention is a new PyTorch API that combines the flexibility of PyTorch with the performance of optimized attention methods. It allows researchers to implement various attention mechanisms easily without needing to write custom kernels. Key Points: - Flexibility with Performance: -- Traditional attention methods offer high performance but limited flexibility. -- FlexAttention provides a flexible API that supports many attention variants with minimal effort. - How It Works: -- Standard Attention: Calculates attention scores using queries, keys, and values, normalizes them, and computes the weighted sum of values. -- FlexAttention: Allows for modifications to the attention scores through a flexible scoring function, enabling customization for different types of attention mechanisms. - Examples of Attention Variants: -- Relative Position Encoding: Adds positional information to attention scores based on relative positions. -- ALiBi Bias: Applies a bias to attention scores to account for position differences, enhancing model performance. -- Causal Mask: Masks future positions in the sequence to ensure the model only attends to previous positions, useful for autoregressive models. Handling Sparsity: -- BlockMask: Efficiently manages sparse attention masks, such as those used in causal attention, by creating masks that prevent unnecessary computations. Learn More: https://lnkd.in/g58pxB4A

8

Master Advanced Information Retrieval: Cutting-edge Techniques to Optimize the Selection of Relevant Documents with Langchain to Create Excellent RAGs #llm #llmops #chstgpt4 #gemini #mlops #aiml https://lnkd.in/dSsjgsX5

10

https://lnkd.in/gidruDdb Probably all of you already know, but I still want to share some exciting developments in neural network architecture! KAN networks, leveraging the Kolmogorov-Arnold representation theorem, present a compelling alternative to MLPs. With KAN, inputs undergo matrix multiplication with multiple nodes, fostering more expressive representations. Additionally, KAN integrates activation functions into weights, revolutionizing the traditional approach. Furthermore, the use of Bezier curves for approximation and the introduction of learned activation functions exemplify the diverse capabilities of KAN networks. Despite its potential, KAN's slower training efficiency and scalability concerns call for further investigation and optimization. Share your thoughts on the potential of KAN networks vis-a-vis MLPs. How do you envision addressing the efficiency and scalability challenges? #KANnetworks #MLPs #NeuralNetworks

1

It's not every day that I get to talk about something that I think will change the future of analytics. Today is one of those days. Introducing: Snowflake Native Amplitude One of the top requests we get from data leaders is to be able to leverage all the investment they've made in their Data Cloud with Snowflake by having Amplitude run directly on top of the data. Today we are able to deliver on that promise. Snowflake Native Amplitude is the same Amplitude Analytics our users know and love, completely replatformed on Snowflake. No ETL, no multiple copies of the data, no issues with data drift or mutability. Amplitude querying directly on top of Snowflake via SQL and returning the results in our point and click dashboards. Curtis Liu and our engineering team spent the last 3 years creating a version of Amplitude Analytics that works with SQL natively on top of Data Warehouses. This required a massive engineering effort. We rebuilt all of our chart querying capabilities, from segmentation, to funnels, to retention, to journeys, to event properties, to every other chart type on top of the data warehouse. We needed to ensure essential features like user timelines, complex segments, microscope, and cohorts worked effectively in the same way as they did on our existing custom Nova database. We then spent a huge amount of time optimizing the hell out of the performance. (Data Warehouses are SLOW if you don't tune them correctly, particularly on critical reports like funnels and user journeys.) As a result, Amplitude is now the first fully replatformed enterprise app on Snowflake. It just works. Sign up for free and start using it right away! We also believe in this so much that we are going big right out of the gate with the launch. Anyone can use full Amplitude analytics, up to 5 seats, completely free of charge. Data leaders have been begging me for this functionality from Amplitude for years. We are excited to finally deliver it today. Read more (and signup!) on Snowflake Native Amplitude here: https://lnkd.in/gbvPDy3P

795

82 Kommentare