Clementine Yang
1 followers · 913 views
on the atlas — 32
- 2312.069423 savers
- Immutable object - Wikipedia1 savers
- Transformers Are Secretly Collectives of Spin Systems | mcbal1 savers
- Hopfield network7 savers
- CUDA C++ Programming Guide5 savers
- CUTLASS: Fast Linear Algebra in CUDA C++ | NVIDIA Technical Blog2 savers
- How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklog24 savers
- 2306.099831 savers
- 2311.085761 savers
- 2212.038271 savers
- 2310.014052 savers
- 1512.044551 savers
- 2201.017701 savers
- 2312.006881 savers
- pdf1 savers
- The 37 Implementation Details of Proximal Policy Optimization · The ICLR Blog Track1 savers
- Growing Action Spaces1 savers
- reinforcement learning - How does one stack multiple observations in the input layer of a convolutional neural network? - Artificial Intelligence Stack Exchange1 savers
- 2203.005431 savers
- 2006.107421 savers
- EPFL_TH5203.pdf1 savers
- untitled1 savers
- Extended Data Table 2 Simulation parameters for actuator, sensor and current diffusion models1 savers
- Extended Data Fig. 5: Training progress. | Nature1 savers
- Towards practical reinforcement learning for tokamak magnetic control - ScienceDirect1 savers
- NgHaradaRussell-shaping-ICML1999.pdf1 savers
- Proximal Policy Optimization Algorithms | alphaXiv1 savers
- 2312.108121 savers
- Scenario trajectory optimization and control on STEP - ScienceDirect1 savers
- Magnetic control of tokamak plasmas through deep reinforcement learning | Nature1 savers
- IEEE Xplore Full-Text PDF:36 savers
- 1301.37814 savers
highlights — 219
By default, we use the last token representation in this paper.
2310.01405Thus, a '\n' 'The' 'amount' 'of' 'happiness' 'in' 'the' 'scenario' 'is' ' ' 0 5 10 15 20 25 30 35 Layer Test Accuracy 55 60 65 70 75 80 Figure 5: The representation at the con- cept token “happiness” in middle layers and the representation at the last token in middle and later layers yield high ac- curacy on the utility estimation task. natural position for collecting neural activity associated with concepts is the tokens corresponding to in the task template T c defined in step 1.
2310.01405The scenarios with high utility evoke distinct neural trajectories within the representation space compared to those with lower utility.
2310.01405These findings are incompatible with a purely circuit-based account of cognition and are more closely aligned with the Hopfieldian view in cognitive neuroscience
2310.01405These findings are incompatible with a purely circuit-based account of cognition and are more closely aligned with the Hopfieldian view in cognitive neuroscience
2310.01405two metrics, usefulness and safety
2312.06942the red-team instructs GPT-4 to look for good opportunities to insert backdoors and to occasionally take such opportunities when they appea
2312.06942The blue team (which is standing in for the developers of an AI system) proposes a strategy we call trusted monitoring , where we monitor GPT-4’s outputs with a trusted but less capable model, GPT-3.5.
2312.06942We want to be robust to intentional subversion from GPT-4 (but we aren’t worried about GPT-3.5)
2312.06942. Our paper builds on this long line of work by demonstrating that many safety-relevant concepts and processes appear to emerge in LLM representations, enabling us to directly monitor and control these aspects of model cognition via representation engineering.
2310.01405not limited to text models.
2310.01405emergent, semantically meaningful internal structure
2310.01405sgemm_naive >>(M, N, K, alpha, A, B, beta, C);
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a WorklogFor our first kernel, we’ll use the grid, block and thread hierarchy to assign each thread a unique entry in the result matrix C
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklogor program performance, as we’ll see later, it’s not a good idea to treat all threads in the same block as equals.
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklogthe thread hierarchy we just talked about mostly concerns program correctness
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a WorklogThreads that are in the same block have access to the same shared memory region (SMEM).
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a WorklogEach block consists of up to 1024 individual threads
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a WorklogEach invocation of a CUDA kernel creates a new grid, which consists of multiple blocks
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklogcomputation is ordered in a three-level hierarchy
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a WorklogNote: Assume equal scaling and units for all signals in the plot, if signal scaling or units are not explicitly indicated.
Scenario trajectory optimization and control on STEP - ScienceDirectA loosely defined precipice for this decision seems to be that for 𝑓 𝐺 𝑊 ≳ 0 . 5 the optimizer will tend toward dominant bootstrap current drive in the high 𝑓 𝐺 𝑊 region and for 𝑓 𝐺 𝑊 ≤ 0 . 5 , the optimizer will choose dominant ECHCD in the low 𝑓 𝐺 𝑊 region
Scenario trajectory optimization and control on STEP - ScienceDirectwhere 𝑒 𝑖 is the difference between the plasma current measurement and reference values at time 𝑖 .
Towards practical reinforcement learning for tokamak magnetic control - ScienceDirectwho train mixtures of two policies with an actor-critic approach, learning a single value function for the current mixture of policies. The mixture contains a policy that may be harder to learn but has a higher perfor- mance ceiling
Growing Action SpacesHowever, these methods cannot directly be applied to grow the action space in the model-free setting
Growing Action SpacesSimilar curric- ula also appear often in hierarchical reinforcement learning, where skills can be learned in comparatively easy settings and then composed in more complex ways later
Growing Action SpacesIn a self-supervised control setting, Murali et al. (2018) use sensitivity analysis to automatically define a curriculum over action dimensions and prioritise their search space.
Growing Action Spacesand provides low variance updates
Growing Action Spacesestimate optimal value functions for multi- ple action spaces simultaneousl
Growing Action Spacescurriculum of progressively growing action spaces
Growing Action SpacesDuring training, the executed control policy is stochastic to explore successful control options. This stochastic policy is represented by a diagonal Gaussian distribution over coil actions.
Magnetic control of tokamak plasmas through deep reinforcement learning | Naturehis was also done in the original DQN paper. This is done mainly to provide more context, i.e. to partially solve the "partial observability" problem
reinforcement learning - How does one stack multiple observations in the input layer of a convolutional neural network? - Artificial Intelligence Stack Exchangeby iteratively updating three components in turn: a policy π (in this case SAC), an encoder φ , and a dynamics model ˆ P
2006.10742batch Monte Carlo setting for simplicity
2203.00543provided a theoretical characterisation of how a given representation affects generalization in reinforcement learning
2203.00543using the same architecture as ours but exchange our bisimulation loss with (1) a reconstruction loss (“ Reconstruction ”) and (2) contrastive predictive coding
2006.10742we evaluate our method in a clean setting without distractors, as well as a much more difficult setting with distractors. We compare against several baseline
2006.10742substantially more robust to task-irrelevant distractors
2006.10742the bisimulation metric is the reward difference added to the Wasserstein distance between transition distributions:
2006.10742Exact partitioning with bisimulation relations is generally impractical in continuous state spaces, as the relation is highly sensitive to infinitesimal changes in the reward function or dynamics. For this reason, Bisimulation Metrics (Ferns et al., 2011; Ferns & Precup, 2014; Castro, 2020) softens the concept of state partitions, and instead defines a pseudometric space ( S ,d ) , where a distance function d : S×S 7→ R ≥ 0 measures the “behavioral similarity” between two state
2006.10742two states are bisimilar if they share both the same immediate reward and equivalent distributions over the next bisimilar states
2006.10742directly learning a representation space with the properties of bisimulation metrics
2006.10742pproximately equivalent state-action pairs are aggregated
2006.10742is substantially more robust to complex distractors
2006.10742by directly learning a bisimilarity-based represen- tation, we guarantee a representation that generates a Lipschitz MDP.
2006.10742our objective directly learns a representation where distance in latent space is the bisimulation metric
2006.10742often still require significant manual tuning
2006.10742jointly train the encoder and the dynamics model end-to-end
2006.10742n practice, there is no guarantee that the learned representation will capture useful information for the control task,
2006.10742when compared to prior approaches that use reconstruction losses or contrastive losses
2006.10742