Jennifer Martay

Quick Start Instructions: build & use car within 10 mins. Info to learn from code also. Good Christmas present. NOTE: need separate #Microbit! #LearnByDoing #STEM Using HolaSmart HS1002 Robot car for Micro:bit -- code install + using t... youtu.be/xy2msiZkkD0?si… via @YouTube

Building #RobotCar from #HolaSmart 7 year old built this on his own in 5 minutes - and is very excited to use it (once we upload codes to #Microbit)! Great Christmas present #STEM @KeyestudioRobot HolaSmart Robot Car for Micro:bit HS1002 youtu.be/5lPKqIjWUlE?si… via @YouTube

@laurencemarks64 Oh, I agree - was just putting my 2 cents in :)

@j_martay I’m sure you do.. 😊. It’s just that there are way more people in academia who don’t.

I'm forced to conclude that the difference between working in Industry and academia is that in industry you can send somebody an email with the expectation that you might get a reply. At some point..

Outstanding lecture series on #NeuralNetworks! Very clear, logical progression, coding and theory - freely available to all. Thank you, @VizuaraAI! #NNFS #AI 20 minutes summary: Building, training and testing neural networks from ... youtu.be/QkFUUfLxd4o?si… via @YouTube

Lectures 28-31 covered -Training v Validation v Testing -#Generalization v #Overfitting w Solutions of: +L1/L2 Regularization +Dropout layers @VizuaraAI #NNFS #NeuralNetworks Lecture 31 - Dropout layers in neural networks - Full code youtu.be/tkvlspCbLqo?si… via @YouTube

@laurencemarks64 Excellent example of multi tasking :)

@j_martay A while... 3 or 4 hours I'd say.. It solved whilst I was at the pub quiz on 4 cores.. When I've done the actual metal forming with the 3d printed tooling I'll post some proper metrics. Unless it doesn't work!

Also this week seeing if you can use 3D printed plastic tooling to form a coil bracket for a BSA Bantam.. looks like you can. Watch this space..

@laurencemarks64 That's an incredible video and solution! About how long does it take to solve?

Lectures 22–27: Progression of optimizers for #NeuralNetworks -Gradient Descent -Learning Rate Decay -Momentum -AdaGrad -RMSProp -Adam +Clear explanations w videos! @VizuaraAI #NNFS Lecture 27 -Coding the ADAM optimizer for neural networks youtu.be/7e0nX6WpnDM?si… via @YouTube

Lectures 10–21: Theory & coding for forward passes and #backpropagation in #NeuralNetworks -Gradients wrt weights, biases, and inputs in fully connected layers -Gradients for ReLU activation and combined Softmax + Cross-Entropy Loss #NNFS youtu.be/sPlBSyJeDqA?si… via @YouTube

Very cool - might try this with my kids!

#Optimization - from random weights to improving loss The thought process is great. "What if we tried this? No, this doesn't work because... What about this? Yes, but..." Really well done, @VizuaraAI #NNFS #NeuralNetworks #AI youtu.be/WVFMxpTD09w?si… via @YouTube

45-min session on: -Categorical Cross Entropy Loss, w only targets or one-hot encoded -Accuracy -#Python Coding Loss and accuracy made a lot of sense - needed to re-watch the coding. Such a clever use of logs! #NNFS #NeuralNetworks #ML #AI youtu.be/QgfINhYPnH0?si… via @YouTube

@OOEC_Oxford It really is a great pub - so nice to have it nearby!

It was very nice to be back at The Butcher's Arms in Headington yesterday evening - we definitely shouldn't leave it so long next time. We are lucky to have such a good pub on the doorstep #Fullers

Get overview of #NeuralNetworks w Codemy video -- then do @VizuaraAI's series Pro: Teacher repeats it's all simple - which you start to believe! Con: Cliffhanger: we create but don't use a NN (next steps behind pay wall). #AI #STEM youtu.be/JHWqWIoac2I?si… via @YouTube

Lecture 5: sum & broadcast arrays=good for #programming, pointless? for #NeuralNetworks? Lecture 6: #ReLU & #softmax #ActivationFunctions. Now I get why Lecture 5 was taught! Thanks for your wonderful #Educational material, @VizuaraAI! #AI youtu.be/SP372QpruDg?si… via @YouTube

@BiomechPhil @Deku007 @Sensors_MDPI Congratulations!

New paper by newly minted PhD @Deku007 using IMUs to predict marker positions for biomechanical modelling mdpi.com/3498308 @Sensors_MDPI

Refresher on object-oriented programming! #OOP Created random-number weight and bias matrices for DenseLayer in #NeuralNetwork. Explained topic to #ChatGPT. It pointed out inaccuracies/unclear wording. youtu.be/qFMpq46hEjE?si… via @YouTube youtu.be/qFMpq46hEjE?si… via @YouTube

In today's CPD, we added layers to forward pass #NeuralNetworks Topic would have fit well in previous lesson, but perhaps @VizuaraAI wanted to pace material for viewers. I may need a 2nd series to learn the theory. #NumPy #AIForBeginners youtu.be/NxwM_eI0Qvk?si… via @YouTube

Lecture uses dot product to code: -Neuron -Layer of neurons -Batch of data into layer Lecture focuses on math rather than theory of #NeuralNetworks - and left me w questions like "Why do different batches have different weights?" #AI #NumPy youtu.be/mK_PfqM88OY?si… via @YouTube