Charles H. Martin, PhD’s Post

I've been curious for a while about accurately finding similarity between two time series. While I know Euclidean distance can work, it doesn’t always cut it for time series data, especially when things get a little more complex. I came across the Dynamic Time Warping (DTW) algorithm, read a bunch of papers, and found that it works really well for comparing time series. It gives a solid metric for finding the most similar sequence. That said, I still haven’t completely figured out the best way to match time series, but DTW is an amazing technique. Additionally, I learned about cross-correlation and compared its performance with DTW. DTW has higher sensitivity for detecting small differences between two time series than cross-correlation, although both techniques have similar computational times. - DTW handles shifts and distortions in time, making it much more flexible than traditional methods like Euclidean distance. - It can compare time series of different lengths by aligning them optimally. - Higher sensitivity to subtle differences compared to cross-correlation. DTW is computationally expensive, it would be interesting learn if there is any other better technique for comparing two time series. #research #learning

📢 Late Advertising of Our Previous AISTATS Work! 📢 Excited to share that our paper, "Optimal Estimation of Gaussian (Poly)trees," presented at AISTATS 2024, is now available online! Dive into the details here: https://lnkd.in/gd5WTsmH. In this work, we develop optimal algorithms for learning Gaussian trees from data. We tackle both distribution learning (in KL distance) and structure learning (exact recovery). Key highlights: - The first approach utilizes the Chow-Liu algorithm for efficient learning of an optimal tree-structured distribution. - The second approach modifies the PC algorithm for polytrees, employing partial correlation as a conditional independence tester for constraint-based structure learning. - We provide explicit finite-sample guarantees for both methods and demonstrate their optimality by deriving matching lower bounds. Our findings establish the optimal sample complexity for learning Gaussian trees. Additionally, we achieved nearly optimal testers for mutual information and conditional mutual information between Gaussian variables. Check out our work and explore the exciting advancements in this area! #AISTATS2024 #GaussianTrees #DataScience #MachineLearning #Research #OptimalSampleComplexity

A simplified overview of the concept of gradient descent https://lnkd.in/e8n5bwKx

Learning About K-Nearest Neighbors (KNN) and Its Hyperparameters Recently explored the K-Nearest Neighbors (KNN) algorithm, a simple yet powerful classification and regression method. KNN predicts a target based on the majority class or average of the 'K' nearest data points. While effective, getting the best out of KNN requires tuning a few key hyperparameters: 1️⃣ K (Number of Neighbors): Determines how many neighbors influence the prediction. A small K makes the model sensitive to noise, while a larger K smooths the decision boundary. 2️⃣ Distance Metric: Defines how distance is measured (e.g., Euclidean, Manhattan). The choice affects how "neighborliness" is calculated, impacting model performance. 3️⃣ Weighting of Neighbors: Allows closer neighbors to have more influence than those farther away. Options like "uniform" (equal weight) or "distance" (weights based on proximity) can improve results. Tuning these parameters can make a significant difference in KNN’s effectiveness. It’s fascinating how a simple algorithm can be so adaptable! #MachineLearning #KNN #Hyperparameters #DataScience

https://lnkd.in/g_vTNknY This was known that randomised algorithm like SGD or similar optimizers do things in order of size of dataset. 🤔 but the randomised SVD ... 🤔 interesting ... do check out. this is a good intro ... #machinelearning

Exploring Mixture Models with EM and Nelder-Mead Algorithms 🧠 We often rely on the Expectation-Maximization (EM) algorithm for Gaussian Mixture Models (GMMs), and for good reason – it's efficient and widely used! But in this brief study, I wanted to highlight that EM isn't the only option. While EM excels in many scenarios, the Nelder-Mead Simplex method also performs well in mixture modeling. ⚙️ Check out the code below! 👇 https://lnkd.in/e2QDFJ_8 I’m excited to explore further how these algorithms can be applied to real-world data and different types of distributions. Always learning, always optimizing! 🔧 #DataScience #MachineLearning #MixtureModels #EMAlgorithm #NelderMead #GaussianMixtureModels #Optimization #ParameterEstimation

Graph inference is a fascinating topic, and a powerful force multiplier for enhancing the value of any knowledge graph. For Plangs! I’ve chosen to defer more sophisticated graph inference to stay focused on delivering a solid 1.0 release. I’m eager to dive deeper into refining this feature post-launch. For now, just some notes on the topic I added to the repo. #KnowledgeGraphs #GraphInference Source: https://lnkd.in/e6SF7UTV Check the draft at https://eoga.dev

The next installment of our graph theory series is here. This time, it's about planar graphs:

Restart!! Today was #Day-01 of LEVEL UP Series. Today's Topics: 1. DSA - Graphs - 3 hrs -Topological Sort. - Kahn's algorithm. - Dijkstra's algorithm. 2. Aptitude - 2.5 hrs. 3. OS - 3 hrs Time spent : 8 hrs 30 min.

Wrapped up 𝐃𝐞𝐜𝐢𝐬𝐢𝐨𝐧 𝐓𝐫𝐞𝐞𝐬 𝐚𝐧𝐝 𝐄𝐧𝐬𝐞𝐦𝐛𝐥𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠( bagging, boosting, and random forests). Notes: Decision Trees: https://lnkd.in/dBuTrY8p Ensemble learning: https://lnkd.in/dQt_HhP3 Coded Decision Trees and Random Forests from scratch: https://lnkd.in/dTxR2fB9 Onto Dimensionality Reduction...