- My take of SORA
- Resume from the middle of an epoch with pytorch-lightning(in one line of code)
- Reading notes: Information-theoretic analysis of generalizationcapability of learning algorithms
- Reading notes: unsupervised word translation paper: Word translation without parallel data
- Some coding work: faster-rcnn under and captioning codebase under pytorch
- Reading notes: trainable greedy decoding
- No label? Make one! Unsupervised Learning by Predicting Noise
- Summary for Structured Prediction Cascades
- ICLR 2017 summary
- Reading notes: Triplets from different parents, CycleGAN, DiscoGAN, DualGAN
- Reading notes: Progressive NN + Pathnet + EWC
- My first attempt on Pytorch
- Reading notes: RL with unsupervised auxiliary tasks
- Reading notes: Using Reinforcement Leanring Algorithm to Generate Image captions
- Some interesting things happened when building SpatialAdaptiveAveragePooling
- Reading notes: Pixel CNN, Wavenet, Language modeling with GCNN (Brief)
- Reading notes: Third person imitation learning
- Neuraltalk2 in tensorflow I started to learn torch by neuraltalk2. And now I start to learn tensorflow by creating a neuraltalk2.
- Detection using Densecap Densecap provides a similar framework of faster-rcnn which however produces captions for each region. I modify it to make it a faster-rcnn.
- Autocomplete using RNN: trained on arxiv data. I had this idea actually, hoping to assist me to write paper. (My poor English!). Glad someone did this.
- Share some more codes in Torch.
- Learn from fb.resnet.torch: Another textbook way to load data I have discussed loading data in neuraltalk2. This time is learning from fb.resnet.torch.
- Cong to AlphaGo: Let's learn torch from Torch based projects on github Here is some repositories I collected on github which are implemented in torch/Lua. These can be helpful for us to get used to torch.
- Available ResNets and neuraltalk2 using ResNet The ResNet was finally released in caffe, however I could not use it. But thank facebook, now I can.
- Some thoughts about AlphaGo I want to talk briefly about this 'exciting' powerful Go AI.
- Learn from neuraltalk2: Wow, why my network model file is so big! I have just met the problem seeing my saved network model is extremely big. And the answer can be found in the neuraltalk2.
- Learn from neuraltalk2: A textbook way to load data I have discussed in serveral blogs talking about memory, loading data etc. This is the best I have ever seen.
- Exceed Lua Heap Limit As I mentioned before, there is the annoying Lua Heap Limit. This is some solution.
- Terrible torch json and the bridge between Lua and Python I was trying to pass dict from python to Torch using json, but json in Torch works bad.
- How to deal with GPU 'out of memory' To be honest, it's not a tutorial to save gpu memory usage. It's more like a discussion. So feel free to comment.
- Learn torch from a COOL application: using DL to imitate artistic style It's a super cool stuff, trying to transform a normal picture to a given artistic picture style. And you can try it by yourself and meanwhile learn torch.
- Start with torch For newbies who first get hands on deep learning and torch.
-
Start a new blog using Jekyll
Here is the new blog. I really hope I can manage it quite well.
subscribe via RSS