Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Splits training

HungBerlinGuy's profile picture

HungBerlinGuy

114,076 subscribers

34,572 Aufrufe • vor 1 Jahr •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

10 Kommentare

Profilbild von thickhungbottom
thickhungbottomvor 1 Jahr

I love splits training 😍

Profilbild von carter
cartervor 1 Jahr

So hot

Profilbild von HornyBiMen88
HornyBiMen88vor 1 Jahr

Geil , wenn ich in diese gym wäre - würde ich jedes mal deine Ladung schlucken

Profilbild von 🍆 #bulginbulges (68k) 📦
🍆 #bulginbulges (68k) 📦vor 1 Jahr

Uff, da macht Trainieren besonders Spaß. Also nachher in der Dusche 🔥

Profilbild von sk8erboy82
sk8erboy82vor 1 Jahr

Da würde icj jetzt mit trainieren 🔥🔥🔥🥵

Profilbild von Master mo
Master movor 1 Jahr

Geil

Profilbild von IG: official_mcenzlee_kash
IG: official_mcenzlee_kashvor 1 Jahr

@sircummings_ii thought this was your for a moment ☺️

Profilbild von UkrainianGayInMunich 🇺🇦🇩🇪(18+)
UkrainianGayInMunich 🇺🇦🇩🇪(18+)vor 1 Jahr

Ach, es muss sehr lecker sein 🥵🥰😍

Profilbild von Pedro Reis Ribeiro
Pedro Reis Ribeirovor 1 Jahr

I just 💗 that dick HEAD

Profilbild von GymbiLondon
GymbiLondonvor 1 Jahr

We need to workout together

Ähnliche Videos

Introducing Beginner Workout Splits — the complete training guide built for total beginners, whatever equipment you've got.
0:48

Introducing Beginner Workout Splits — the complete training guide built for total beginners, whatever equipment you've got.

weightlossmessiah

77,519 Aufrufe • vor 2 Monaten

This is actually a VERY GOOD Training Splits Tier List by everyone’s favorite influencer: Jeff Nippard And what he says at the end is definitely what matters most in real life!
0:58

This is actually a VERY GOOD Training Splits Tier List by everyone’s favorite influencer: Jeff Nippard And what he says at the end is definitely what matters most in real life!

Dean Turner

176,935 Aufrufe • vor 8 Monaten

Unwrapped by @NikoKinbaku
0:24

Unwrapped by @NikoKinbaku

HungBerlinGuy

20,633 Aufrufe • vor 2 Jahren

There is NO DEBATE…. 1. Upper/Lower/OFF/Upper/Lower/OFF/OFF 2. Upper/Lower/OFF/Pull/Push/Lower/OFF These are the BEST training splits known to man We will be sure to update you if this ever changes!
0:31

There is NO DEBATE…. 1. Upper/Lower/OFF/Upper/Lower/OFF/OFF 2. Upper/Lower/OFF/Pull/Push/Lower/OFF These are the BEST training splits known to man We will be sure to update you if this ever changes!

Dean Turner

40,640 Aufrufe • vor 1 Monat

Training a muscle once per week is leaving gains on the table. Muscle protein synthesis stays elevated 24-48 hours after training. Then the muscle starts breaking down. Training Monday means growth stimulus until Wednesday. Then atrophy until next Monday. You could easily end the week with nothing to show for it. Train each muscle 2-3x weekly: Hit it Monday, Wednesday, Friday. Keep protein synthesis elevated constantly. More frequent training = more total weekly stimulus = more growth. This is basic biology but bro splits ignore it completely.
0:23

Training a muscle once per week is leaving gains on the table. Muscle protein synthesis stays elevated 24-48 hours after training. Then the muscle starts breaking down. Training Monday means growth stimulus until Wednesday. Then atrophy until next Monday. You could easily end the week with nothing to show for it. Train each muscle 2-3x weekly: Hit it Monday, Wednesday, Friday. Keep protein synthesis elevated constantly. More frequent training = more total weekly stimulus = more growth. This is basic biology but bro splits ignore it completely.

Sama Hoole

297,688 Aufrufe • vor 8 Monaten

Sometimes looking a little sus is exactly what your hips need. This month’s Field Note is breaking down 3 of my favourite ways to implement this type of hip work into your training so you can build qualities that carry over to positions like the pancake splits. Link Below 👇🏼
0:28

Sometimes looking a little sus is exactly what your hips need. This month’s Field Note is breaking down 3 of my favourite ways to implement this type of hip work into your training so you can build qualities that carry over to positions like the pancake splits. Link Below 👇🏼

Lucas Hardie

81,094 Aufrufe • vor 1 Monat

I was 15 years old when I realized Bro Splits are absolutely silly Can they produce results? Sure Will they produce the best results all else equal? Not even CLOSE Training a muscle group 1x every 7 days = TONS of progress left on the table How do people STILL not get this?
0:26

I was 15 years old when I realized Bro Splits are absolutely silly Can they produce results? Sure Will they produce the best results all else equal? Not even CLOSE Training a muscle group 1x every 7 days = TONS of progress left on the table How do people STILL not get this?

Dean Turner

144,672 Aufrufe • vor 5 Monaten

Open hips = better movement, stronger lines, fewer limits. Do these drills before skills that demand mobility, or after training to decompress. Breakdancing, acrobatics, yoga—doesn’t matter. If your hips are tight, your moves will suffer. These hip openers directly support handstands, splits, airflares, flares, windmills. Stay upright, control the pelvis, drive the knees out… and yes—embrace the burn.
1:14

Open hips = better movement, stronger lines, fewer limits. Do these drills before skills that demand mobility, or after training to decompress. Breakdancing, acrobatics, yoga—doesn’t matter. If your hips are tight, your moves will suffer. These hip openers directly support handstands, splits, airflares, flares, windmills. Stay upright, control the pelvis, drive the knees out… and yes—embrace the burn.

ᴀʀᴛ ᴏꜰ ᴘʜʏꜱɪQᴜᴇ

146,722 Aufrufe • vor 9 Monaten

Reasoning From Scratch round 8, the finale! This last lecture in the series covers distillation as an efficient alternative to the previous reinforcement learning training runs. (Planning to take a brief break, but I will be back!) 00:00 Introduction and series recap 01:07 Why distill a reasoning model? 02:51 Soft vs. hard distillation 11:06 On-policy vs. off-policy distillation 18:09 Preparing data for reasoning distillation 22:35 Loading the DeepSeek R1 MATH dataset 28:04 Inspecting reasoning traces and final answers 31:27 Tokenization and chat templates 35:58 Pre-tokenizing the training dataset 39:01 Sequence lengths and the 2,048-token limit 42:38 Creating training and validation splits 45:32 Loading the Qwen3-0.6B base model 46:24 Computing answer-token log probabilities 54:10 Cross-entropy loss for distillation 57:25 Training and validation loss helpers 1:02:03 Implementing the distillation training loop 1:07:26 Running the standalone training script 1:09:41 Plotting training and validation losses 1:11:31 Evaluating checkpoints on MATH-500 1:14:09 Comparing teachers and distillation results 1:16:34 Future directions for reasoning models 1:19:41 Bonus exercises and next steps
1:21:17

Reasoning From Scratch round 8, the finale! This last lecture in the series covers distillation as an efficient alternative to the previous reinforcement learning training runs. (Planning to take a brief break, but I will be back!) 00:00 Introduction and series recap 01:07 Why distill a reasoning model? 02:51 Soft vs. hard distillation 11:06 On-policy vs. off-policy distillation 18:09 Preparing data for reasoning distillation 22:35 Loading the DeepSeek R1 MATH dataset 28:04 Inspecting reasoning traces and final answers 31:27 Tokenization and chat templates 35:58 Pre-tokenizing the training dataset 39:01 Sequence lengths and the 2,048-token limit 42:38 Creating training and validation splits 45:32 Loading the Qwen3-0.6B base model 46:24 Computing answer-token log probabilities 54:10 Cross-entropy loss for distillation 57:25 Training and validation loss helpers 1:02:03 Implementing the distillation training loop 1:07:26 Running the standalone training script 1:09:41 Plotting training and validation losses 1:11:31 Evaluating checkpoints on MATH-500 1:14:09 Comparing teachers and distillation results 1:16:34 Future directions for reasoning models 1:19:41 Bonus exercises and next steps

Sebastian Raschka

32,493 Aufrufe • vor 2 Tagen

Ready to Level Up 📊 your 🦾Training? Here’s 3 Plyos to improve muscular coordination! Muscular Coordination 🧠⚡️ training teaches your nervous system and muscles to 🔥fire in sync. This reduces wasted energy, optimizes your stride mechanics, and translates raw strength into explosive 🧨 Power💥 RLE Jumps- Knee to Hip “Puncture your Belt” Pogos- Dorsi Flex you ankles “Hit the ceiling, Floor is hot coals” Split Box Jumps- Get whole of foot on both boxes “create big splits quick” Bonus: Double Scissor Jumps- Switch legs fast “Create big shapes” Bonus: Timed Wickets- Het hips under abs “Strike the ground” With muscular coordination think big shapes, strike through the ground (make it rumble), while getting off the ground super quick!
0:26

Ready to Level Up 📊 your 🦾Training? Here’s 3 Plyos to improve muscular coordination! Muscular Coordination 🧠⚡️ training teaches your nervous system and muscles to 🔥fire in sync. This reduces wasted energy, optimizes your stride mechanics, and translates raw strength into explosive 🧨 Power💥 RLE Jumps- Knee to Hip “Puncture your Belt” Pogos- Dorsi Flex you ankles “Hit the ceiling, Floor is hot coals” Split Box Jumps- Get whole of foot on both boxes “create big splits quick” Bonus: Double Scissor Jumps- Switch legs fast “Create big shapes” Bonus: Timed Wickets- Het hips under abs “Strike the ground” With muscular coordination think big shapes, strike through the ground (make it rumble), while getting off the ground super quick!

Coach A Vogel

12,347 Aufrufe • vor 3 Monaten

A masterclass on Google's TPU v8 Networking. Two TPU chips? Pssh. We already knew workload-specific silicon was here. But two scale-up networking topologies? That's the actual Google TPU news. Workload-specific interconnects. Think about that. New Semi Doped with Vikram Sekar and Austin Lyons. Copper? Yep. Optics? Yep. What we cover: - TPU splits in two: 8t training, 8i inference. - Virgo: 47 Pb/s scale-out fabric, 100% OCS. - Boardfly scale-up: copper PCB + AECs inside racks, OCS between groups. 16 hops → 7. - Training uses 3D torus (Rubik's Cube). - Inference doesn't. Workload-specific topologies now. - Dragonfly traces to a 2008 paper by Kim, Dally, Scott, Abts. Abts went on to build Groq's interconnect before Nvidia. Chapters: 0:00 Intro 0:21 Two TPUs for two workloads 2:31 HBM, SRAM, and Axion CPUs 7:22 Why networking is the new bottleneck 17:14 Virgo: rebuilding scale-out on optics 25:24 3D torus Rubik's Cube scale-up for training 34:50 Boardfly: scale-up for MoE inference 42:07 Workload-specific everything $GOOGL
46:59

A masterclass on Google's TPU v8 Networking. Two TPU chips? Pssh. We already knew workload-specific silicon was here. But two scale-up networking topologies? That's the actual Google TPU news. Workload-specific interconnects. Think about that. New Semi Doped with Vikram Sekar and Austin Lyons. Copper? Yep. Optics? Yep. What we cover: - TPU splits in two: 8t training, 8i inference. - Virgo: 47 Pb/s scale-out fabric, 100% OCS. - Boardfly scale-up: copper PCB + AECs inside racks, OCS between groups. 16 hops → 7. - Training uses 3D torus (Rubik's Cube). - Inference doesn't. Workload-specific topologies now. - Dragonfly traces to a 2008 paper by Kim, Dally, Scott, Abts. Abts went on to build Groq's interconnect before Nvidia. Chapters: 0:00 Intro 0:21 Two TPUs for two workloads 2:31 HBM, SRAM, and Axion CPUs 7:22 Why networking is the new bottleneck 17:14 Virgo: rebuilding scale-out on optics 25:24 3D torus Rubik's Cube scale-up for training 34:50 Boardfly: scale-up for MoE inference 42:07 Workload-specific everything $GOOGL

Semi Doped

93,264 Aufrufe • vor 5 Monaten

The new Huberman Lab episode is out: Build Muscle, Great Posture & Resilience to Injury | Jeff Cavaliere 0:00 Jeff Cavaliere 2:43 Lower Back, Back Pain 10:06 Tool: Exercises for Lower Back Pain & Strengthen Glutes 15:29 Sponsors: David & Our Place 18:05 Walking Exercises to Strengthen Glutes 23:23 Small Focused Exercises & Timing; Workout Soreness & Pain 27:08 Tools: "Old Man" Test, Functional Strength Tests 35:08 Sports, Movement Imbalance 40:57 Tool: Training Like An Athlete 46:44 Sponsor: AG1 48:28 Inner Elbow Pain, Tool: Grip Modification 54:21 Shoulder, Rotator Cuff Training, Tool: External Rotation Exercise 1:06:50 Tool: Neck Training; Women, Posture 1:15:20 Longevity, Strength & Agility; Pain & Construction Zone Analogy 1:19:50 Sponsor: Joovv 1:21:12 Cardio, Bike, Jumping Rope; Fat Loss: Zone 2 or HIIT? 1:29:27 Nutrition, Tools: Calorie Counting; Plate Method 1:40:18 Foot Stability, Tool: Foot Strength Test; Longevity 1:48:15 Sponsor: Function 1:49:53 Warm-Up Sets; Reps in Reserve or Train to Failure?; Work Sets 2:00:30 Training Frequency; Tool: Real-Life Constraints, Split the Splits 2:13:58 Zero-Cost Support, YouTube, Spotify & Apple Follow, Reviews & Feedback, Sponsors, Protocols Book, Social Media, Neural Network Newsletter Includes paid partnerships.
2:16:49

The new Huberman Lab episode is out: Build Muscle, Great Posture & Resilience to Injury | Jeff Cavaliere 0:00 Jeff Cavaliere 2:43 Lower Back, Back Pain 10:06 Tool: Exercises for Lower Back Pain & Strengthen Glutes 15:29 Sponsors: David & Our Place 18:05 Walking Exercises to Strengthen Glutes 23:23 Small Focused Exercises & Timing; Workout Soreness & Pain 27:08 Tools: "Old Man" Test, Functional Strength Tests 35:08 Sports, Movement Imbalance 40:57 Tool: Training Like An Athlete 46:44 Sponsor: AG1 48:28 Inner Elbow Pain, Tool: Grip Modification 54:21 Shoulder, Rotator Cuff Training, Tool: External Rotation Exercise 1:06:50 Tool: Neck Training; Women, Posture 1:15:20 Longevity, Strength & Agility; Pain & Construction Zone Analogy 1:19:50 Sponsor: Joovv 1:21:12 Cardio, Bike, Jumping Rope; Fat Loss: Zone 2 or HIIT? 1:29:27 Nutrition, Tools: Calorie Counting; Plate Method 1:40:18 Foot Stability, Tool: Foot Strength Test; Longevity 1:48:15 Sponsor: Function 1:49:53 Warm-Up Sets; Reps in Reserve or Train to Failure?; Work Sets 2:00:30 Training Frequency; Tool: Real-Life Constraints, Split the Splits 2:13:58 Zero-Cost Support, YouTube, Spotify & Apple Follow, Reviews & Feedback, Sponsors, Protocols Book, Social Media, Neural Network Newsletter Includes paid partnerships.

Andrew D. Huberman, Ph.D.

703,234 Aufrufe • vor 4 Monaten

Introducing FlyOCR 🪰 - I trained a fly brain to read a PDF It uses the full MaleCNS v1.0 fruit fly connectcome. The architecture is inspired by doomfly by The fly splits a pdf image into individual glyphs, maps pixels into receptor activations, runs simplified current-based dynamics across the 166k neurons and 25m edges in the circuit, applies a compact readout model on the downstream spikes, and concatenates everything into the parsed output. On reading an actual Microsoft 10-k, the fly gets ~86% over the balance sheet heading, but is largely able to read the numeric values correctly. Over 1.7k+ sampled glyphs (chars+digits) it gets 87% accuracy. With enough training it might match some of the latter-generation MNIST models! Maybe eventually we’ll replace our doc parsing VLMs with flies. Full video below. Repo with full code + report:
0:40

Introducing FlyOCR 🪰 - I trained a fly brain to read a PDF It uses the full MaleCNS v1.0 fruit fly connectcome. The architecture is inspired by doomfly by The fly splits a pdf image into individual glyphs, maps pixels into receptor activations, runs simplified current-based dynamics across the 166k neurons and 25m edges in the circuit, applies a compact readout model on the downstream spikes, and concatenates everything into the parsed output. On reading an actual Microsoft 10-k, the fly gets ~86% over the balance sheet heading, but is largely able to read the numeric values correctly. Over 1.7k+ sampled glyphs (chars+digits) it gets 87% accuracy. With enough training it might match some of the latter-generation MNIST models! Maybe eventually we’ll replace our doc parsing VLMs with flies. Full video below. Repo with full code + report:

Jerry Liu

38,858 Aufrufe • vor 1 Monat

Covenant Labs just did a 90-minute AMA breaking down their 3 Bittensor subnets. templar. basilica. grail. Pre-training, compute, and post-training under one roof. Most people missed it. Here's everything they said. Covenant is building what they call the "end to end intelligence continuum." Three subnets. Three layers of the AI stack. All permissionless. Templar (SN3) handles decentralized pre-training. Basilica (SN39) handles compute. Grail (SN81) handles RL post-training. Sam Dare, the lead, put it bluntly. Decentralized training is "humanity's last dance." Not about beating OpenAI head to head. About creating optionality. About making it cheap enough for anyone to train models. The gap between academia and frontier labs is growing exponentially. Researchers can't afford to experiment. The actual training run costs 5% of the reported budget. The other 95% is experimentation. If Covenant cracks cheap training, that entire surface area opens up. On Templar specifically: • Hit 39% emission on Bittensor. Highest since Apex was the only subnet on the network • Covenant-72B trained permissionlessly with 70+ contributors on commodity internet • 1.1 trillion tokens processed. No centralized data center • Performance competitive with LLaMA-2-70B On Grail, something flew under the radar. They built Pulse. A weight synchronization method that compresses model updates by 100x. • In RL post-training, only ~1% of weights update per step • Pulse exploits that sparsity. Lossless compression • Prime Intellect's comparable system took 14 minutes to sync a 30B model • Pulse makes decentralized RL training actually feasible at scale • Already used by Cursor The lead researcher on Grail said they've trained on math, code, and GPU kernels. Got 40-60% improvement on benchmarks. Working toward agentic training with 100K+ token context and 30B+ parameter models. On Basilica, the compute subnet: The team was blunt. Just reselling GPU hours is a 5-10% margin game. Traditional compute providers already do that. Their play is value-added services. • "GPU as code." No dashboard. No UI. Agents interact via SDK • Custom scheduler that places workloads across heterogeneous hardware • Verification checks for GPU, CPU, bandwidth, memory, storage, and OS security • Partnerships with providers like Mass Compute for 10-20% below market pricing • Miners compete on useful infrastructure, not just GPU hours Sam then went on a rant about the miner burn debate. His take: Bittensor had to grow up. dTAO introduced investors. The old "miners are God" philosophy doesn't hold. • Subnet owners have a duty to protect token value • Miners are a resource optimization exercise, not a cost reduction exercise • 100% miner emissions on compute subnets = immediate sell pressure • The 41% miner allocation is arbitrary. Different business models need different splits • Fish (who started burns) agreed. Burns usually mean the validation isn't mature enough The bigger point. You can't police burns. Subnets just send to their own keys instead of the burn address. Subnet 28 does exactly that. Sam's position: judge subnets on outcomes, not process. Const has changed the protocol 9-10 times in 2 years. That iteration speed is Bittensor's actual moat. The whole Covenant thesis is playing out in real time. TAO is up 100%+ in a month. Jensen Huang name-dropped the network. Grayscale has an ETF filing. But the real story is three subnets quietly building every layer of decentralized AI.
1:39:04

Covenant Labs just did a 90-minute AMA breaking down their 3 Bittensor subnets. templar. basilica. grail. Pre-training, compute, and post-training under one roof. Most people missed it. Here's everything they said. Covenant is building what they call the "end to end intelligence continuum." Three subnets. Three layers of the AI stack. All permissionless. Templar (SN3) handles decentralized pre-training. Basilica (SN39) handles compute. Grail (SN81) handles RL post-training. Sam Dare, the lead, put it bluntly. Decentralized training is "humanity's last dance." Not about beating OpenAI head to head. About creating optionality. About making it cheap enough for anyone to train models. The gap between academia and frontier labs is growing exponentially. Researchers can't afford to experiment. The actual training run costs 5% of the reported budget. The other 95% is experimentation. If Covenant cracks cheap training, that entire surface area opens up. On Templar specifically: • Hit 39% emission on Bittensor. Highest since Apex was the only subnet on the network • Covenant-72B trained permissionlessly with 70+ contributors on commodity internet • 1.1 trillion tokens processed. No centralized data center • Performance competitive with LLaMA-2-70B On Grail, something flew under the radar. They built Pulse. A weight synchronization method that compresses model updates by 100x. • In RL post-training, only ~1% of weights update per step • Pulse exploits that sparsity. Lossless compression • Prime Intellect's comparable system took 14 minutes to sync a 30B model • Pulse makes decentralized RL training actually feasible at scale • Already used by Cursor The lead researcher on Grail said they've trained on math, code, and GPU kernels. Got 40-60% improvement on benchmarks. Working toward agentic training with 100K+ token context and 30B+ parameter models. On Basilica, the compute subnet: The team was blunt. Just reselling GPU hours is a 5-10% margin game. Traditional compute providers already do that. Their play is value-added services. • "GPU as code." No dashboard. No UI. Agents interact via SDK • Custom scheduler that places workloads across heterogeneous hardware • Verification checks for GPU, CPU, bandwidth, memory, storage, and OS security • Partnerships with providers like Mass Compute for 10-20% below market pricing • Miners compete on useful infrastructure, not just GPU hours Sam then went on a rant about the miner burn debate. His take: Bittensor had to grow up. dTAO introduced investors. The old "miners are God" philosophy doesn't hold. • Subnet owners have a duty to protect token value • Miners are a resource optimization exercise, not a cost reduction exercise • 100% miner emissions on compute subnets = immediate sell pressure • The 41% miner allocation is arbitrary. Different business models need different splits • Fish (who started burns) agreed. Burns usually mean the validation isn't mature enough The bigger point. You can't police burns. Subnets just send to their own keys instead of the burn address. Subnet 28 does exactly that. Sam's position: judge subnets on outcomes, not process. Const has changed the protocol 9-10 times in 2 years. That iteration speed is Bittensor's actual moat. The whole Covenant thesis is playing out in real time. TAO is up 100%+ in a month. Jensen Huang name-dropped the network. Grayscale has an ETF filing. But the real story is three subnets quietly building every layer of decentralized AI.

Jesus Martinez

26,763 Aufrufe • vor 6 Monaten

F*CK. SpaceXAI gave away an entire day of Grok Bot training for free. People still spend thousands on agent courses. This 35-minute workshop shows how SpaceXAI moves from a broken product funnel to a verified PR with a team of bots. step 1 → assign the roles: chief of staff, data scientist, product manager, designer, and engineering manager step 2 → the data scientist analyzes the funnel, catches the human mistake, and finds the real drop-off step 3 → the product manager turns that finding into a PRD with P0, P1, and P2 requirements step 4 → the designer creates two Figma options using the company's design system and hands the winner to engineering step 5 → the engineering manager splits the work across five engineers, launches cloud agents, and routes the output through verification SpaceXAI says Grok Bot already accounts for a double-digit percentage of its merged PRs. One data question. One PRD. Two designs. Five engineers. One PR. Watch it today, then read the article on building your own Grok Bot team from scratch below.
35:00

F*CK. SpaceXAI gave away an entire day of Grok Bot training for free. People still spend thousands on agent courses. This 35-minute workshop shows how SpaceXAI moves from a broken product funnel to a verified PR with a team of bots. step 1 → assign the roles: chief of staff, data scientist, product manager, designer, and engineering manager step 2 → the data scientist analyzes the funnel, catches the human mistake, and finds the real drop-off step 3 → the product manager turns that finding into a PRD with P0, P1, and P2 requirements step 4 → the designer creates two Figma options using the company's design system and hands the winner to engineering step 5 → the engineering manager splits the work across five engineers, launches cloud agents, and routes the output through verification SpaceXAI says Grok Bot already accounts for a double-digit percentage of its merged PRs. One data question. One PRD. Two designs. Five engineers. One PR. Watch it today, then read the article on building your own Grok Bot team from scratch below.

GenAI

19,535 Aufrufe • vor 21 Tagen

SVM by hand ✍️ ~ 19 steps walkthrough below (Linear vs RBF) Support Vector Machines reigned supreme in machine learning before the deep learning revolution. An SVM predicts with dot products, the same matrix multiplication every model uses. What it does not do is train by backpropagation: it is fitted by convex optimization, so there is no matrix-multiplication backward pass for a GPU to accelerate. I drew and calculated two SVMs by hand: a linear one (top) and an RBF one (bottom), classifying the same two test vectors. Goal: turn six training vectors and their learned coefficients into a prediction, and see what changing the kernel actually changes. = 1. Given = Six training vectors, their labels, and the coefficients and bias already learned. A coefficient of zero means that vector is not a support vector: too far from the boundary to matter. = 2. Linear kernel, test vector 1 = Let us take the dot product of the test vector with every training vector. The dot product stands in for cosine similarity, and the column of results is the first column of the kernel matrix K. = 3. Linear kernel, test vector 2 = We do the same for the second, and K is complete. = 4. Signed weights = Let us multiply each coefficient by its label. The second training vector drops out here, because its coefficient is 0. = 5. Weighted combination = We multiply the signed weights through K and add the bias b. The result is a signed distance to the decision boundary: 17 and 5. = 6. Classify = Let us take the sign. Both are positive. = 7 to 11. RBF kernel, test vector 1 = Now the same picture with a different kernel, in five moves: square the differences, sum them, take the square root for the L2 distance, multiply by minus gamma, and raise e to that power. The negation is what turns a distance into a similarity, and gamma controls how far a single training vector's influence reaches. = 12 to 16. RBF kernel, test vector 2 = We repeat all five. The numbers change, the moves do not. = 17 to 19. Decision boundary, again = Signed weights, weighted combination, sign. Identical arithmetic to steps 4 through 6, on a K that was built a completely different way. The outputs: Linear K, first column = [13, 25, 12, 15, 19, 27] Linear decision values = 17 and 5, both positive RBF decision values = -2 and 1, so negative and positive The takeaway: the kernel is the only thing that changed, and it changed the answer. The linear SVM calls both test vectors positive; the RBF one splits them. Everything after the kernel matrix, the signed weights and the weighted combination and the sign, is the same page of arithmetic twice. 💾 Save this post!
0:44

SVM by hand ✍️ ~ 19 steps walkthrough below (Linear vs RBF) Support Vector Machines reigned supreme in machine learning before the deep learning revolution. An SVM predicts with dot products, the same matrix multiplication every model uses. What it does not do is train by backpropagation: it is fitted by convex optimization, so there is no matrix-multiplication backward pass for a GPU to accelerate. I drew and calculated two SVMs by hand: a linear one (top) and an RBF one (bottom), classifying the same two test vectors. Goal: turn six training vectors and their learned coefficients into a prediction, and see what changing the kernel actually changes. = 1. Given = Six training vectors, their labels, and the coefficients and bias already learned. A coefficient of zero means that vector is not a support vector: too far from the boundary to matter. = 2. Linear kernel, test vector 1 = Let us take the dot product of the test vector with every training vector. The dot product stands in for cosine similarity, and the column of results is the first column of the kernel matrix K. = 3. Linear kernel, test vector 2 = We do the same for the second, and K is complete. = 4. Signed weights = Let us multiply each coefficient by its label. The second training vector drops out here, because its coefficient is 0. = 5. Weighted combination = We multiply the signed weights through K and add the bias b. The result is a signed distance to the decision boundary: 17 and 5. = 6. Classify = Let us take the sign. Both are positive. = 7 to 11. RBF kernel, test vector 1 = Now the same picture with a different kernel, in five moves: square the differences, sum them, take the square root for the L2 distance, multiply by minus gamma, and raise e to that power. The negation is what turns a distance into a similarity, and gamma controls how far a single training vector's influence reaches. = 12 to 16. RBF kernel, test vector 2 = We repeat all five. The numbers change, the moves do not. = 17 to 19. Decision boundary, again = Signed weights, weighted combination, sign. Identical arithmetic to steps 4 through 6, on a K that was built a completely different way. The outputs: Linear K, first column = [13, 25, 12, 15, 19, 27] Linear decision values = 17 and 5, both positive RBF decision values = -2 and 1, so negative and positive The takeaway: the kernel is the only thing that changed, and it changed the answer. The linear SVM calls both test vectors positive; the RBF one splits them. Everything after the kernel matrix, the signed weights and the weighted combination and the sign, is the same page of arithmetic twice. 💾 Save this post!

Tom Yeh

16,916 Aufrufe • vor 2 Monaten