Find resources and get questions answered, A place to discuss PyTorch code, issues, install, research, Discover, publish, and reuse pre-trained models. Mathematically, if you have a vector valued function d.backward() For policies applicable to the PyTorch Project a Series of LF Projects, LLC, The accuracy of the model is calculated on the test data and shows the percentage of the right prediction. why the grad is changed, what the backward function do? Making statements based on opinion; back them up with references or personal experience. Now all parameters in the model, except the parameters of model.fc, are frozen. In the given direction of filter, the gradient image defines its intensity from each pixel of the original image and the pixels with large gradient values become possible edge pixels. Therefore, a convolution layer with 64 channels and kernel size of 3 x 3 would detect 64 distinct features, each of size 3 x 3. backward function is the implement of BP(back propagation), What is torch.mean(w1) for? Without further ado, let's get started! Asking the user for input until they give a valid response, Minimising the environmental effects of my dyson brain. Join the PyTorch developer community to contribute, learn, and get your questions answered. How do I print colored text to the terminal? At this point, you have everything you need to train your neural network. The image gradient can be computed on tensors and the edges are constructed on PyTorch platform and you can refer the code as follows. gradient is a tensor of the same shape as Q, and it represents the If you do not provide this information, your Next, we run the input data through the model through each of its layers to make a prediction. By default w1.grad 0.6667 = 2/3 = 0.333 * 2. No, really. To run the project, click the Start Debugging button on the toolbar, or press F5. the parameters using gradient descent. \frac{\partial y_{m}}{\partial x_{1}} & \cdots & \frac{\partial y_{m}}{\partial x_{n}} you can change the shape, size and operations at every iteration if So,dy/dx_i = 1/N, where N is the element number of x. During the training process, the network will process the input through all the layers, compute the loss to understand how far the predicted label of the image is falling from the correct one, and propagate the gradients back into the network to update the weights of the layers. How do I check whether a file exists without exceptions? The implementation follows the 1-step finite difference method as followed For this example, we load a pretrained resnet18 model from torchvision. If spacing is a scalar then If you enjoyed this article, please recommend it and share it! What is the point of Thrower's Bandolier? torch.gradient(input, *, spacing=1, dim=None, edge_order=1) List of Tensors Estimates the gradient of a function g : \mathbb {R}^n \rightarrow \mathbb {R} g: Rn R in one or more dimensions using the second-order accurate central differences method. Is there a proper earth ground point in this switch box? Label in pretrained models has I need to use the gradient maps as loss functions for back propagation to update network parameters, like TV Loss used in style transfer. What can a lawyer do if the client wants him to be acquitted of everything despite serious evidence? In a NN, parameters that dont compute gradients are usually called frozen parameters. The number of out-channels in the layer serves as the number of in-channels to the next layer. Lets walk through a small example to demonstrate this. For example: A Convolution layer with in-channels=3, out-channels=10, and kernel-size=6 will get the RGB image (3 channels) as an input, and it will apply 10 feature detectors to the images with the kernel size of 6x6. By default, when spacing is not This is because sobel_h finds horizontal edges, which are discovered by the derivative in the y direction. accurate if ggg is in C3C^3C3 (it has at least 3 continuous derivatives), and the estimation can be The PyTorch Foundation supports the PyTorch open source Gx is the gradient approximation for vertical changes and Gy is the horizontal gradient approximation. how to compute the gradient of an image in pytorch. The leaf nodes in blue represent our leaf tensors a and b. DAGs are dynamic in PyTorch Interested in learning more about neural network with PyTorch? y = mean(x) = 1/N * \sum x_i \vdots & \ddots & \vdots\\ you can also use kornia.spatial_gradient to compute gradients of an image. YES Have you updated the Stable-Diffusion-WebUI to the latest version? If you mean gradient of each perceptron of each layer then model [0].weight.grad will show you exactly that (for 1st layer). In PyTorch, the neural network package contains various loss functions that form the building blocks of deep neural networks. For tensors that dont require The images have to be loaded in to a range of [0, 1] and then normalized using mean = [0.485, 0.456, 0.406] and std = [0.229, 0.224, 0.225]. The most recognized utilization of image gradient is edge detection that based on convolving the image with a filter. To train the model, you have to loop over our data iterator, feed the inputs to the network, and optimize. the coordinates are (t0[1], t1[2], t2[3]), dim (int, list of int, optional) the dimension or dimensions to approximate the gradient over. If you preorder a special airline meal (e.g. Your numbers won't be exactly the same - trianing depends on many factors, and won't always return identifical results - but they should look similar. autograd then: computes the gradients from each .grad_fn, accumulates them in the respective tensors .grad attribute, and. Sign in Loss function gives us the understanding of how well a model behaves after each iteration of optimization on the training set. Every technique has its own python file (e.g. Check out my LinkedIn profile. Refresh the. Disconnect between goals and daily tasksIs it me, or the industry? Staging Ground Beta 1 Recap, and Reviewers needed for Beta 2. \end{array}\right)\], # check if collected gradients are correct, # Freeze all the parameters in the network, Deep Learning with PyTorch: A 60 Minute Blitz, Visualizing Models, Data, and Training with TensorBoard, TorchVision Object Detection Finetuning Tutorial, Transfer Learning for Computer Vision Tutorial, Optimizing Vision Transformer Model for Deployment, Language Modeling with nn.Transformer and TorchText, Fast Transformer Inference with Better Transformer, NLP From Scratch: Classifying Names with a Character-Level RNN, NLP From Scratch: Generating Names with a Character-Level RNN, NLP From Scratch: Translation with a Sequence to Sequence Network and Attention, Text classification with the torchtext library, Real Time Inference on Raspberry Pi 4 (30 fps! w.r.t. (here is 0.6667 0.6667 0.6667) from torchvision import transforms img (Tensor) An (N, C, H, W) input tensor where C is the number of image channels, Tuple of (dy, dx) with each gradient of shape [N, C, H, W]. I have some problem with getting the output gradient of input. \end{array}\right)\], \[\vec{v} When spacing is specified, it modifies the relationship between input and input coordinates. the indices are multiplied by the scalar to produce the coordinates. how the input tensors indices relate to sample coordinates. 1-element tensor) or with gradient w.r.t. respect to \(\vec{x}\) is a Jacobian matrix \(J\): Generally speaking, torch.autograd is an engine for computing The main objective is to reduce the loss function's value by changing the weight vector values through backpropagation in neural networks. If spacing is a list of scalars then the corresponding They are considered as Weak. You signed in with another tab or window. from torch.autograd import Variable Short story taking place on a toroidal planet or moon involving flying. Simple add the run the code below: Now that we have a classification model, the next step is to convert the model to the ONNX format, More info about Internet Explorer and Microsoft Edge. How to match a specific column position till the end of line? Or is there a better option? It is useful to freeze part of your model if you know in advance that you wont need the gradients of those parameters In your answer the gradients are swapped. You defined h_x and w_x, however you do not use these in the defined function. Saliency Map. They're most commonly used in computer vision applications. In resnet, the classifier is the last linear layer model.fc. I need to compute the gradient (dx, dy) of an image, so how to do it in pytroch? here is a reference code (I am not sure can it be for computing the gradient of an image ) import torch from torch.autograd import Variable w1 = Variable (torch.Tensor ( [1.0,2.0,3.0]),requires_grad=True) tensor([[ 0.3333, 0.5000, 1.0000, 1.3333], # The following example is a replication of the previous one with explicit, second-order accurate central differences method. the variable, As you can see above, we've a tensor filled with 20's, so average them would return 20. The values are organized such that the gradient of \end{array}\right)\left(\begin{array}{c} = The console window will pop up and will be able to see the process of training. \vdots\\ Find centralized, trusted content and collaborate around the technologies you use most. How to follow the signal when reading the schematic? functions to make this guess. In this section, you will get a conceptual By clicking Post Your Answer, you agree to our terms of service, privacy policy and cookie policy. - Satya Prakash Dash May 30, 2021 at 3:36 What you mention is parameter gradient I think (taking y = wx + b parameter gradient is w and b here)? (tensor([[ 4.5000, 9.0000, 18.0000, 36.0000]. By clicking Post Your Answer, you agree to our terms of service, privacy policy and cookie policy. Here is a small example: For web site terms of use, trademark policy and other policies applicable to The PyTorch Foundation please see Letting xxx be an interior point and x+hrx+h_rx+hr be point neighboring it, the partial gradient at X=P(G) To analyze traffic and optimize your experience, we serve cookies on this site. In this tutorial, you will use a Classification loss function based on Define the loss function with Classification Cross-Entropy loss and an Adam Optimizer. \frac{\partial y_{1}}{\partial x_{1}} & \cdots & \frac{\partial y_{m}}{\partial x_{1}}\\ # indices and input coordinates changes based on dimension. and stores them in the respective tensors .grad attribute. f(x+hr)f(x+h_r)f(x+hr) is estimated using: where xrx_rxr is a number in the interval [x,x+hr][x, x+ h_r][x,x+hr] and using the fact that fC3f \in C^3fC3 gradient computation DAG. Note that when dim is specified the elements of It is simple mnist model. Implementing Custom Loss Functions in PyTorch. We use the models prediction and the corresponding label to calculate the error (loss). import torch here is a reference code (I am not sure can it be for computing the gradient of an image ) Making statements based on opinion; back them up with references or personal experience. This is why you got 0.333 in the grad. Notice although we register all the parameters in the optimizer, backward() do the BP work automatically, thanks for the autograd mechanism of PyTorch. By iterating over a huge dataset of inputs, the network will learn to set its weights to achieve the best results. Gradients are now deposited in a.grad and b.grad. Finally, lets add the main code. As usual, the operations we learnt previously for tensors apply for tensors with gradients. Revision 825d17f3. \(J^{T}\cdot \vec{v}\). # For example, below, the indices of the innermost dimension 0, 1, 2, 3 translate, # to coordinates of [0, 3, 6, 9], and the indices of the outermost dimension. Low-Weakand Weak-Highthresholds: we set the pixels with high intensity to 1, the pixels with Low intensity to 0 and between the two thresholds we set them to 0.5. Mathematically, the value at each interior point of a partial derivative print(w1.grad) PyTorch datasets allow us to specify one or more transformation functions which are applied to the images as they are loaded. How do I combine a background-image and CSS3 gradient on the same element? The accuracy of the model is calculated on the test data and shows the percentage of the right prediction. graph (DAG) consisting of This will will initiate model training, save the model, and display the results on the screen. (tensor([[ 1.0000, 1.5000, 3.0000, 4.0000], # When spacing is a list of scalars, the relationship between the tensor. The below sections detail the workings of autograd - feel free to skip them. If you have found these useful in your research, presentations, school work, projects or workshops, feel free to cite using this DOI. Then, we used PyTorch to build our VGG-16 model from scratch along with understanding different types of layers available in torch. [-1, -2, -1]]), b = b.view((1,1,3,3)) P=transforms.Compose([transforms.ToPILImage()]), ten=torch.unbind(T(img)) proportionate to the error in its guess. img = Image.open(/home/soumya/Downloads/PhotographicImageSynthesis_master/result_256p/final/frankfurt_000000_000294_gtFine_color.png.jpg).convert(LA) Tensors with Gradients Creating Tensors with Gradients Allows accumulation of gradients Method 1: Create tensor with gradients gradients, setting this attribute to False excludes it from the Or do I have the reason for my issue completely wrong to begin with? Perceptual Evaluation of Speech Quality (PESQ), Scale-Invariant Signal-to-Distortion Ratio (SI-SDR), Scale-Invariant Signal-to-Noise Ratio (SI-SNR), Short-Time Objective Intelligibility (STOI), Error Relative Global Dim. If a law is new but its interpretation is vague, can the courts directly ask the drafters the intent and official interpretation of their law? rev2023.3.3.43278. Find resources and get questions answered, A place to discuss PyTorch code, issues, install, research, Discover, publish, and reuse pre-trained models, Click here Loss value is different from model accuracy. Our network will be structured with the following 14 layers: Conv -> BatchNorm -> ReLU -> Conv -> BatchNorm -> ReLU -> MaxPool -> Conv -> BatchNorm -> ReLU -> Conv -> BatchNorm -> ReLU -> Linear. PyTorch doesnt have a dedicated library for GPU use, but you can manually define the execution device. This package contains modules, extensible classes and all the required components to build neural networks. How should I do it? The PyTorch Foundation is a project of The Linux Foundation. The optimizer adjusts each parameter by its gradient stored in .grad. For example, below the indices of the innermost, # 0, 1, 2, 3 translate to coordinates of [0, 2, 4, 6], and the indices of. NVIDIA GeForce GTX 1660, If the issue is specific to an error while training, please provide a screenshot of training parameters or the Access comprehensive developer documentation for PyTorch, Get in-depth tutorials for beginners and advanced developers, Find development resources and get your questions answered. In this tutorial we will cover PyTorch hooks and how to use them to debug our backward pass, visualise activations and modify gradients. - Allows calculation of gradients w.r.t. Now, it's time to put that data to use. the only parameters that are computing gradients (and hence updated in gradient descent) To extract the feature representations more precisely we can compute the image gradient to the edge constructions of a given image. Sign up for a free GitHub account to open an issue and contact its maintainers and the community. Dreambooth revision is 5075d4845243fac5607bc4cd448f86c64d6168df Diffusers version is *0.14.0* Torch version is 1.13.1+cu117 Torch vision version 0.14.1+cu117, Have you read the Readme? 3 Likes How to remove the border highlight on an input text element. Why is this sentence from The Great Gatsby grammatical? As you defined, the loss value will be printed every 1,000 batches of images or five times for every iteration over the training set. As before, we load a pretrained resnet18 model, and freeze all the parameters. The nodes represent the backward functions Each node of the computation graph, with the exception of leaf nodes, can be considered as a function which takes some inputs and produces an output. How to check the output gradient by each layer in pytorch in my code? Can I tell police to wait and call a lawyer when served with a search warrant? maybe this question is a little stupid, any help appreciated! You can run the code for this section in this jupyter notebook link. Synthesis (ERGAS), Learned Perceptual Image Patch Similarity (LPIPS), Structural Similarity Index Measure (SSIM), Symmetric Mean Absolute Percentage Error (SMAPE). The following other layers are involved in our network: The CNN is a feed-forward network. YES In PyTorch, the neural network package contains various loss functions that form the building blocks of deep neural networks. If you do not provide this information, your issue will be automatically closed. requires_grad=True. . Powered by Discourse, best viewed with JavaScript enabled, http://pytorch.org/docs/0.3.0/torch.html?highlight=torch%20mean#torch.mean. Let S is the source image and there are two 3 x 3 sobel kernels Sx and Sy to compute the approximations of gradient in the direction of vertical and horizontal directions respectively. \end{array}\right)=\left(\begin{array}{c} A tensor without gradients just for comparison. PyTorch image classification with pre-trained networks; PyTorch object detection with pre-trained networks; By the end of this guide, you will have learned: . You can see the kernel used by the sobel_h operator is taking the derivative in the y direction. The basic principle is: hi! May I ask what the purpose of h_x and w_x are? Both are computed as, Where * represents the 2D convolution operation. J. Rafid Siddiqui, PhD. I need to compute the gradient(dx, dy) of an image, so how to do it in pytroch? By clicking or navigating, you agree to allow our usage of cookies. Read PyTorch Lightning's Privacy Policy. operations (along with the resulting new tensors) in a directed acyclic [I(x+1, y)-[I(x, y)]] are at the (x, y) location. specified, the samples are entirely described by input, and the mapping of input coordinates = The value of each partial derivative at the boundary points is computed differently. to write down an expression for what the gradient should be. How Intuit democratizes AI development across teams through reusability. The lower it is, the slower the training will be. Lets take a look at a single training step. Connect and share knowledge within a single location that is structured and easy to search. Feel free to try divisions, mean or standard deviation! Testing with the batch of images, the model got right 7 images from the batch of 10. \frac{\partial \bf{y}}{\partial x_{1}} & And similarly to access the gradients of the first layer model[0].weight.grad and model[0].bias.grad will be the gradients. backwards from the output, collecting the derivatives of the error with So model[0].weight and model[0].bias are the weights and biases of the first layer. tensor([[ 1.0000, 1.5000, 3.0000, 4.0000], # A scalar value for spacing modifies the relationship between tensor indices, # and input coordinates by multiplying the indices to find the, # coordinates. Site design / logo 2023 Stack Exchange Inc; user contributions licensed under CC BY-SA. If a law is new but its interpretation is vague, can the courts directly ask the drafters the intent and official interpretation of their law? x=ten[0].unsqueeze(0).unsqueeze(0), a=np.array([[1, 0, -1],[2,0,-2],[1,0,-1]]) By clicking Sign up for GitHub, you agree to our terms of service and Function My Name is Anumol, an engineering post graduate. By tracing this graph from roots to leaves, you can And There is a question how to check the output gradient by each layer in my code. Why does Mister Mxyzptlk need to have a weakness in the comics? That is, given any vector \(\vec{v}\), compute the product We need to explicitly pass a gradient argument in Q.backward() because it is a vector. one or more dimensions using the second-order accurate central differences method. Have you updated Dreambooth to the latest revision? If you do not do either of the methods above, you'll realize you will get False for checking for gradients. Not the answer you're looking for? It runs the input data through each of its a = torch.Tensor([[1, 0, -1], If you dont clear the gradient, it will add the new gradient to the original. project, which has been established as PyTorch Project a Series of LF Projects, LLC. Describe the bug. what is torch.mean(w1) for? Finally, if spacing is a list of one-dimensional tensors then each tensor specifies the coordinates for If I print model[0].grad after back-propagation, Is it going to be the output gradient by each layer for every epoches? So, what I am trying to understand why I need to divide the 4-D Tensor by tensor(28.) Asking for help, clarification, or responding to other answers. This signals to autograd that every operation on them should be tracked. If you will look at the documentation of torch.nn.Linear here, you will find that there are two variables to this class that you can access. Learn how our community solves real, everyday machine learning problems with PyTorch. # Set the requires_grad_ to the image for retrieving gradients image.requires_grad_() After that, we can catch the gradient by put the . TypeError If img is not of the type Tensor. \frac{\partial l}{\partial x_{1}}\\ .backward() call, autograd starts populating a new graph. Next, we loaded and pre-processed the CIFAR100 dataset using torchvision. The backward pass kicks off when .backward() is called on the DAG (A clear and concise description of what the bug is), What OS? [0, 0, 0], Short story taking place on a toroidal planet or moon involving flying. These functions are defined by parameters The first is: import torch import torch.nn.functional as F def gradient_1order (x,h_x=None,w_x=None): issue will be automatically closed. Learn about PyTorchs features and capabilities. We register all the parameters of the model in the optimizer. Try this: thanks for reply. , My bad, I didn't notice it, sorry for the misunderstanding, I have further edited the answer, How to get the output gradient w.r.t input, discuss.pytorch.org/t/gradients-of-output-w-r-t-input/26905/2, How Intuit democratizes AI development across teams through reusability. Forward Propagation: In forward prop, the NN makes its best guess vision Michael (Michael) March 27, 2017, 5:53pm #1 In my network, I have a output variable A which is of size h w 3, I want to get the gradient of A in the x dimension and y dimension, and calculate their norm as loss function. how to compute the gradient of an image in pytorch. What video game is Charlie playing in Poker Face S01E07? I am training a model on pictures of my faceWhen I start to train my model it charges and gives the following error: OSError: Error no file named diffusion_pytorch_model.bin found in directory C:\ai\stable-diffusion-webui\models\dreambooth[name_of_model]\working. When you create our neural network with PyTorch, you only need to define the forward function. To get the gradient approximation the derivatives of image convolve through the sobel kernels. Awesome, thanks a lot, and what if I would love to know the "output" gradient for each layer? See the documentation here: http://pytorch.org/docs/0.3.0/torch.html?highlight=torch%20mean#torch.mean. conv1=nn.Conv2d(1, 1, kernel_size=3, stride=1, padding=1, bias=False) project, which has been established as PyTorch Project a Series of LF Projects, LLC. By querying the PyTorch Docs, torch.autograd.grad may be useful. How can this new ban on drag possibly be considered constitutional? d = torch.mean(w1) OK How to properly zero your gradient, perform backpropagation, and update your model parameters most deep learning practitioners new to PyTorch make a mistake in this step ; Staging Ground Beta 1 Recap, and Reviewers needed for Beta 2. import torch.nn as nn To analyze traffic and optimize your experience, we serve cookies on this site. w1.grad In NN training, we want gradients of the error They should be edges_y = filters.sobel_h (im) , edges_x = filters.sobel_v (im). \frac{\partial l}{\partial y_{1}}\\ We can use calculus to compute an analytic gradient, i.e. For example, if the indices are (1, 2, 3) and the tensors are (t0, t1, t2), then single input tensor has requires_grad=True. are the weights and bias of the classifier. [1, 0, -1]]), a = a.view((1,1,3,3)) to your account. # doubling the spacing between samples halves the estimated partial gradients. This estimation is Shereese Maynard. Maybe implemented with Convolution 2d filter with require_grad=false (where you set the weights to sobel filters). Next, we load an optimizer, in this case SGD with a learning rate of 0.01 and momentum of 0.9. How do I print colored text to the terminal? Well occasionally send you account related emails. \frac{\partial l}{\partial y_{m}} Manually and Automatically Calculating Gradients Gradients with PyTorch Run Jupyter Notebook You can run the code for this section in this jupyter notebook link. improved by providing closer samples. I am learning to use pytorch (0.4.0) to automate the gradient calculation, however I did not quite understand how to use the backward () and grad, as I'm doing an exercise I need to calculate df / dw using pytorch and making the derivative analytically, returning respectively auto_grad, user_grad, but I did not quite understand the use of This is detailed in the Keyword Arguments section below. The backward function will be automatically defined. They told that we can get the output gradient w.r.t input, I added more explanation, hopefully clearing out any other doubts :), Actually, sample_img.requires_grad = True is included in my code. using the chain rule, propagates all the way to the leaf tensors. objects. Connect and share knowledge within a single location that is structured and easy to search. # Estimates the gradient of f(x)=x^2 at points [-2, -1, 2, 4], # Estimates the gradient of the R^2 -> R function whose samples are, # described by the tensor t. Implicit coordinates are [0, 1] for the outermost, # dimension and [0, 1, 2, 3] for the innermost dimension, and function estimates. external_grad represents \(\vec{v}\). to get the good_gradient Not bad at all and consistent with the model success rate. Learn how our community solves real, everyday machine learning problems with PyTorch. [2, 0, -2], Please find the following lines in the console and paste them below. needed. So firstly when you print the model variable you'll get this output: And if you choose model[0], that means you have selected the first layer of the model. # partial derivative for both dimensions. Let me explain to you! For example, for the operation mean, we have: A forward function computes the value of the loss function, and the backward function computes the gradients of the learnable parameters. res = P(G). # the outermost dimension 0, 1 translate to coordinates of [0, 2]. to be the error. understanding of how autograd helps a neural network train. that acts as our classifier. We could simplify it a bit, since we dont want to compute gradients, but the outputs look great, #Black and white input image x, 1x1xHxW by the TF implementation. Learn more, including about available controls: Cookies Policy. How should I do it? The output tensor of an operation will require gradients even if only a By clicking or navigating, you agree to allow our usage of cookies. If you need to compute the gradient with respect to the input you can do so by calling sample_img.requires_grad_(), or by setting sample_img.requires_grad = True, as suggested in your comments. \left(\begin{array}{ccc}\frac{\partial l}{\partial y_{1}} & \cdots & \frac{\partial l}{\partial y_{m}}\end{array}\right)^{T}\], \[J^{T}\cdot \vec{v}=\left(\begin{array}{ccc} input (Tensor) the tensor that represents the values of the function, spacing (scalar, list of scalar, list of Tensor, optional) spacing can be used to modify We will use a framework called PyTorch to implement this method. Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support. See edge_order below. All pre-trained models expect input images normalized in the same way, i.e. How do I combine a background-image and CSS3 gradient on the same element? good_gradient = torch.ones(*image_shape) / torch.sqrt(image_size) In above the torch.ones(*image_shape) is just filling a 4-D Tensor filled up with 1 and then torch.sqrt(image_size) is just representing the value of tensor(28.) \[y_i\bigr\rvert_{x_i=1} = 5(1 + 1)^2 = 5(2)^2 = 5(4) = 20\], \[\frac{\partial o}{\partial x_i} = \frac{1}{2}[10(x_i+1)]\], \[\frac{\partial o}{\partial x_i}\bigr\rvert_{x_i=1} = \frac{1}{2}[10(1 + 1)] = \frac{10}{2}(2) = 10\], Copyright 2021 Deep Learning Wizard by Ritchie Ng, Manually and Automatically Calculating Gradients, Long Short Term Memory Neural Networks (LSTM), Fully-connected Overcomplete Autoencoder (AE), Forward- and Backward-propagation and Gradient Descent (From Scratch FNN Regression), From Scratch Logistic Regression Classification, Weight Initialization and Activation Functions, Supervised Learning to Reinforcement Learning (RL), Markov Decision Processes (MDP) and Bellman Equations, Fractional Differencing with GPU (GFD), DBS and NVIDIA, September 2019, Deep Learning Introduction, Defence and Science Technology Agency (DSTA) and NVIDIA, June 2019, Oral Presentation for AI for Social Good Workshop ICML, June 2019, IT Youth Leader of The Year 2019, March 2019, AMMI (AIMS) supported by Facebook and Google, November 2018, NExT++ AI in Healthcare and Finance, Nanjing, November 2018, Recap of Facebook PyTorch Developer Conference, San Francisco, September 2018, Facebook PyTorch Developer Conference, San Francisco, September 2018, NUS-MIT-NUHS NVIDIA Image Recognition Workshop, Singapore, July 2018, NVIDIA Self Driving Cars & Healthcare Talk, Singapore, June 2017, NVIDIA Inception Partner Status, Singapore, May 2017.