Patrick Spedding (@spedding) header
Patrick Spedding (@spedding) profile picture

Mathematician | Data scientist | Economics & AI explorer | Teaching maths, building economic data tools & competing in shot put at 60. Still learning.

Sydney, New South Wales
Joined November 2008
Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)Photo by Patrick Spedding (@spedding)
Patrick Spedding retweeted
With the RBA lining up to whack mortgage holders again, here’s the far fairer way to take the heat out of the economy, and tackle inflation. We tax the 1% and rein in their spending on data centres and their obvious excess, don’t keep punching down on mortgage holders.
237
395
36
1,003
25,067
Patrick Spedding retweeted
Backpropagation by hand ✍️ ~ 11 steps walkthrough below Backpropagation is the algorithm that actually trains a neural network, and it is where most people stop following along. It is not calculus you cannot do. It is matrix multiplication, working backward, one layer at a time. So I drew and calculated one entirely by hand. Goal: push the loss gradient back through a 3-layer network and land on a new value for every weight and bias. = 1. Given = A 3-layer perceptron, an input X, predictions Ypred = [0.5, 0.5, 0], and the truth Ytarget = [0, 1, 0]. = 2. Backprop gradient cells = Let us draw empty cells for every gradient we are about to compute. The shape of the answer comes first. = 3. Layer 3 softmax = We get dL/dz3 straight from Ypred minus Ytarget = [0.5, -0.5, 0]. No chain rule needed, and that shortcut is the whole reason softmax and cross-entropy are paired. = 4. Layer 3 weights and biases = Let us multiply dL/dz3 by [a2 | 1]. One multiplication gives the gradient for W3 and b3 together. = 5. Layer 2 activations = We multiply dL/dz3 by W3 to get dL/da2. The gradient moves back across a layer the same way the signal moved forward. = 6. Layer 2 ReLU = Let us pass it through the gate: keep the gradient where the activation was positive, zero it everywhere else. = 7. Layer 2 weights and biases = We multiply dL/dz2 by [a1 | 1]. The same figure as step 4, one layer up. = 8. Layer 1 activations = Let us multiply dL/dz2 by W2. = 9. Layer 1 ReLU = We apply the same gate again, now on a1. = 10. Layer 1 weights and biases = Let us multiply dL/dz1 by [x | 1], and every weight in the network now has a gradient. = 11. Update = We subtract, and the network has learned. In practice a learning rate scales this step. The gradients: dL/dz3 = [0.5, -0.5, 0] dL/da1 = [1, -2, 2, -1] dL/dz1 = [0, -2, 2, -1] The takeaway: matrix multiplication is all you need. Just like the forward pass, backpropagation is matrix multiplications end to end. You can do every one by hand, slowly and imperfectly, which is exactly why a GPU's ability to do them fast mattered so much to deep learning. 💾 Save this post!
189
654
27
4,126
964,108
Patrick Spedding retweeted
Mathematics. Functions. Dance.
49
2,021
136
7,095
326,765
Patrick Spedding retweeted
I figured out a way to save thousands of dollars on ClawdBot By using Opus 4.5 for everything, you are SHREDDING your money and tokens In this video I cover how you can use multiple models to get better performance and save $$ Must watch if you use Clawd:
164
143
16
1,850
159,334
🚀 The Complete Guide to Building AI Agents, From Zero to Production AI Agents are the next big leap in automation, they can think, plan, and execute tasks just like humans 🤖 This guide takes you from basics to building real, production-ready AI Agents — step by step! To get your copy 👇 Like & Repost Comment “AI” Follow (so I can DM you the guide)
1,371
958
31
2,170
237,685
Patrick Spedding retweeted
The strong increase in profits alongside solid revenue growth suggests that the digital world is entering a mature phase, shaped by efficiency, intelligent automation and a more conscious use of technological resources to sustain long-term value. @StatistaCharts via @antgrasso
9
62
3
86
8,222
NSW: "Nearly 2800 NSW teachers left their posts last year in an exodus from the public system" "NSW Teachers Federation president Henry Rajendra said “more needs to be done to bring these numbers down” Source: archive.md/ByxIf
6
22
1
78
2,121
Patrick Spedding retweeted
The beauty of Italy, region by region 🧵 1. Campania: an evening on the Amalfi Coast
654
8,169
628
85,564
14,911,644
Patrick Spedding retweeted
Stanford University just released free online courses. No payment Required. Here are best courses you don't want to miss in 2025:
168
6,625
234
53,385
6,332,715
Patrick Spedding retweeted
What's in store for the Australian economy in 2025? Here's what to look out for this year Both home prices and the share market are tipped to go lower this year. But mortgage holders should finally see some relief. au.finance.yahoo.com/news/wh…
1
10
28
3,446
The one and only Sarah Cooper is back!🤣
302
3,325
236
16,380
1,277,676
Patrick Spedding retweeted
Trump: You have to get out and vote. You won’t have to do it anymore. Four years, it will be fixed, it will be fine. You won’t have to vote anymore.. In four years, you won’t have to vote again.
8,478
17,167
25,491
48,829
81,556,682
Patrick Spedding retweeted
Trump's pattern of denigrating women is deeply disturbing. A selection of egregious comments: - Trump infamously told Billy Bush on Access Hollywood in 2005: "I don't even wait. And when you're a star, they let you do it. You can do anything. Grab 'em by the pussy. You can do anything." - Trump told New York magazine: "Women, you have to treat them like shit." - Trump said of Jessica Leeds, who accused him of sexual assault, "believe me, she would not be my first choice, that I can tell you." - Trump said of Brande Roderick on Celebrity Apprentice "It must be a pretty picture. You dropping to your knees." - Trump told Howard Stern it's "checkout time" for women after the age of 35. - Trump said of Megyn Kelly: "You could see there was blood coming out of her eyes, blood coming out of her... wherever." - Trump tweeted "If Hillary Clinton can't satisfy her husband, what makes her think she can satisfy America?" - Trump attacked Republican rival Carly Fiorina's appearance, saying "Look at that face! Would anyone vote for that?" - Trump stated that if he comes home and "dinner's not ready," he "go[es] through the roof." - Trump called Alicia Machado, the 1996 Miss Universe winner, "Miss Piggy" because she gained weight and "Miss Housekeeping" because of her Hispanic heritage. - Trump said about his daughter Ivanka "She does have a very nice figure. I've said if Ivanka weren't my daughter, perhaps I'd be dating her." - Trump said pregnant employees are an "inconvenience." - In a deposition for the E. Jean Carroll sexual assault lawsuit, Trump claimed that his previous boasts about grabbing women "by the pussy" were historically accurate for "stars" like himself. - Trump bragged to Stern about barging into the Miss Universe dressing rooms while contestants were "standing there with no clothes," saying "I sort of get away with things like that."
Over a mere six-week period, Donald Trump carried out an unprecedented firing spree against government watchdogs: - On April 2, 2020, President Trump fired Intelligence Community Inspector General Michael Atkinson for fulfilling his legal duty to disclose to Congress an "urgent” and credible" whistleblower complaint prompting Republican Senators Richard Burr's rebuke that "an IG must be allowed to conduct...work independent of pressure" and Susan Collins' assertion the dismissal was "not warranted," while Justice Department IG Michael Horowitz lauded Atkinson's "integrity" and "commitment to the rule of law." - On April 7, 2020, President Trump removed Glenn Fine from his position as acting Inspector General of the Department of Defense, claiming without evidence that he was "partisan," despite Fine's long history of nonpartisan service under presidents of both parties, prompting a rare public rebuke from former Defense Secretary Jim Mattis, who praised Fine as "a public servant in the finest tradition of honest, competent governance." - On May 1, 2020, Trump fired Health and Human Services Inspector General Christi Grimm, a career official serving since 1999 praised by her Bush-appointed predecessor for her "highly respected" and "longstanding career of public service," after she released a report based on a survey of 323 hospitals documenting widespread shortages of critical medical supplies amid the COVID-19 pandemic. - On May 15, 2020, President Trump fired State Department Inspector General Steve Linick, who was investigating Secretary of State Mike Pompeo's authorization of an $8 billion arms sale to Saudi Arabia and the UAE without Congressional approval, as well as Pompeo's use of staffers for personal errands, with Trump telling reporters that he had "never even heard of" Linick but fired him after Pompeo “asked me to terminate him." In other words, Trump admitted to firing Linick at the request of the very official he was investigating. #TheCaseAgainstTrump #InspectorGenerals
73
438
39
769
908,142
Patrick Spedding retweeted
MIT University just released free online courses. No payment is required. Here are 7 courses you don't want to miss in 2024:
485
9,528
463
58,560
21,442,694
Patrick Spedding retweeted
Let me try and explain the monumental incompetence of the mismanagement of Australia's gas resources. Vast amounts of gas is exported from Australia, but not by Australians. Its exported by a handful of predominantly foreign owned multinational gas corporations, 🧵👇1/10
109
751
63
1,983
119,775
Patrick Spedding retweeted
Since the Supreme Court's Chevron decision has taken independent regulation powers away from government agencies, let's take this opportunity to review what happens now if Trump wins and the GOP gets to launch Project 2025. Hint: The 2 are connected. 🧵
519
4,661
612
14,273
3,017,091
Patrick Spedding retweeted
Former President Trump failed to impress everyone in a room full of top CEOs Thursday at the Business Roundtable’s quarterly meeting, multiple attendees told CNBC. cnb.cx/45u22ib
136
42
20
104
43,874
Patrick Spedding retweeted
The dumbed-down anti-science Neanderthals are cranking out their deceptively edited clips from the Fauci hearing because clicks and keeping their lemmings as stupid as possible is their primary objective.
147
687
25
3,543
150,693